Journal of Economic Literature 2014, 52(4), 1075–1118
http://dx.doi.org/10.1257/jel.52.4.1075
Behavioral Contract Theory †
Botond Kőszegi*
This review provides a critical survey of psychology-and-economics (“behavioraleconomics”) research in contract theory. First, I introduce the theories of individual
decision making most frequently used in behavioral contract theory, and formally
illustrate some of their implications in contracting settings. Second, I provide a
more comprehensive (but informal) survey of the psychology-and-economics work
on classical contract-theoretic topics: moral hazard, screening, mechanism design,
and incomplete contracts. I also summarize research on a new topic spawned by
psychology and economics, exploitative contracting, that studies contracts designed
primarily to take advantage of agent mistakes. ( JEL A12, D03, D82, D86)
1. Introduction
P
sychology and economics—also known
as behavioral economics—is a mindset for doing economics that espouses the
importance of thinking about the psychological accuracy of models. After a history of identifying deviations from classical
approaches, modeling these deviations formally, and empirically establishing their
importance in economic decisions, the field
* Central European University. I am grateful to Paul
Heidhues and Klaus Schmidt for insightful discussions
and detailed comments on drafts of this review. I am also
hugely indebted to Matthew Rabin for countless conversations over the years that helped clarify my thinking on
many of the issues covered in this paper. I thank Mark
Armstrong, Petra Bischofberger, Andrea Canidio, Takeshi
Murooka, Yuval Salant, Patrick Schmitz, Ferdinand von
Siemens, Rani Spiegler, and four anonymous referees for
comments, and the European Research Council for support under Starting Grant #313341. Kinga Marczell provided excellent research assistance, including drafting the
first version of the auction-theory parts of section 5.
† Go to http://dx.doi.org/10.1257/jel.52.4.1075 to visit
the article page and view author disclosure statement(s).
is in the process of full integration into economic analysis: researchers are using the
new psychologically based models to study
how individual behavior plays out in organizations and markets, what the welfare consequences are, and how policy should respond
to market outcomes—questions in which
economists have always been interested.
This review summarizes and organizes
one important part of the above development, the rapidly growing literature on
behavioral contract theory. In a well-known
textbook less than a decade old, Bolton and
Dewatripont (2005, p. 10) wrote that “even
though there is by now a large literature
exploring a wide range of alternative models
of individual choice . . . there have been relatively few explorations of the implications for
optimal contracting.” By the time of Bolton
and Dewatripont’s book, contract theory had
evolved to a curious point: while researchers had explored optimal contracting in
many complex economic environments, they
largely followed a simplistic approach for
1075
1076
Journal of Economic Literature, Vol. LII (December 2014)
modeling agent behavior. In a moral-hazard
context, for instance, an agent has an outside
option, trades off the monetary benefits and
the costs of effort, and optimally chooses
whether to participate and what to do.
In contrast, recent research has fruitfully
incorporated behavioral-economics ideas into
most of the classical contract-theoretic topics, including moral hazard, screening, auction
theory, and incomplete contracts. In addition,
the recognition that agents often fail to act in
their best interests—and that principals might
know this better than agents do—has inspired
a novel literature on exploitative contracts,
contracts whose exclusive or primary aim is to
take advantage of agent mistakes. For each of
these topics, I attempt to organize the main
insights researchers have uncovered, as well as
point out some of the many additional questions raised by existing work.
Like in any review, it is necessary to draw
some, partly arbitrary, lines regarding which
papers are included. I chose papers that
are “behavioral” in the sense that a formal,
relatively general model with psychological foundations or compelling psychological interpretation is involved, and for which
some evidence supports the specific decisionmaking process. This excludes, for instance,
the very interesting literature on bounded
rationality predicated on optimization with
cognitive, effort, or information-acquisition costs (e.g., Tirole 2009 and Bolton and
Faure-Grimaud 2010), which does not yet
seem to be based solidly on psychology interpretation and evidence.1 Further, I focus on
papers that address design questions related
1 Many papers in this literature, including those cited
above, uncover mechanisms that intuitions suggest would
hold and continue to be important in more psychologically
based models. In other cases, however, the main results
are based on essentially neoclassical optimization principles with novel types of presumed costs. Garicano and Prat
(2013) review some of this work. How the insights in this
literature connect to those based more closely on psychology evidence is a promising topic for future research.
to contracts, mechanisms, or other extended
interactions. This excludes purely gametheoretic issues, such as the winner’s curse
and overbidding in auctions (e.g., Eyster and
Rabin 2005), that are studied in behavioral
game theory; and pure pricing issues, such
as oligopolistic pricing with consumer loss
aversion (e.g., Heidhues and Kőszegi 2008)
that are studied in behavioral industrial
organization.2
I begin in section 2 by presenting the theories of individual decision making in psychology and economics that are most used in
behavioral contract theory, and by working
out some of their implications in simple settings. These theories are (i) loss aversion—
whereby an individual evaluates economic
outcomes relative to reference points, and
weights losses more heavily than similarsized gains; (ii) present bias—whereby an
individual weights an earlier relative to a
delayed outcome more heavily when the
earlier outcome is in the present; (iii) inequity aversion—whereby people in a social
situation dislike advantageous and especially
disadvantageous inequality in material outcomes; and (iv) overconfidence—whereby
a person displays unrealistically positive
beliefs regarding her ability or prospects. In
almost all applications, researchers assume
that the agent (she) behaves according to
one psychologically based model, while the
principal (he) is fully rational and has a classical goal (usually profit maximization).
The main sections of the paper cover work
from psychology and economics on the main
contract-theoretic topics. Deviating from
reviews by Rabin (1998) and DellaVigna
(2009), I organize the discussion p
rimarily
2 See Spiegler (2011) for an excellent textbook coverage of many industrial-organization models. A substantial
amount of research is either at the intersection of behavioral industrial organization and behavioral contract theory,
or is not clearly categorizable, and is hence covered in both
this survey and Spiegler’s textbook; see especially chapters
2–5 and 11 of the book.
Kőszegi: Behavioral Contract Theory
not around behavioral-economics phenomena—whether the underlying model of decision making involves loss aversion, hyperbolic
discounting, etc.—but rather according to
contract-theoretic principles—whether the
interaction qualifies as moral hazard, screening, etc. I do so for two main reasons. First,
this organization reflects the fact that the
literature has been addressing some of the
classical contract-theoretic topics that interest other economists. Second, for some topics, the precise psychological source of the
phenomena in question is not important
(although for other questions it is).
In section 3, I discuss work on moral hazard, where the principal’s main goal in contracting is to provide incentives for the agent
to exert costly effort. Much of this literature
focuses on sources of nonfinancial motivation that affect the agent’s willingness to
work hard. As the most important example,
researchers recognize—following original
discussions by Akerlof (1982) and Akerlof
and Yellen (1988, 1990)—that organizations
are often social in nature, so that agents’
attitudes toward others have an important
impact on behavior. Researchers study the
nature of these and other nonpecuniary
motivations, as well as their interaction with
financial incentives.
In section 4, I summarize existing research
on asymmetric information and screening,
where the principal is attempting to interact with an agent whose private information
he does not know. Psychology and economics introduces a number of important novel
themes, including a new reason for screening contracts: present-biased agents prefer
to write contracts to constrain their own
future behavior, while knowing that information pertaining to optimal behavior will
arrive in the future. In addition, the psychological phenomena emphasized in behavioral
economics also lead to the natural screening
issue of how to deal with agents who exhibit
these phenomena to different extents.
1077
Section 5 turns to a less extensive literature
in behavioral contract theory, mechanism
design—where a principal is interacting with
multiple agents whose private information
he does not know—and its particular application, auction theory. Many of the papers in
this area study the implications of referencedependent utility and loss aversion for auctions. The main results can be categorized
around two forces. First, because loss-averse
bidders dislike risk in how much they will
pay, an auction designer has an incentive
to choose auctions that insure them from
this risk. Second, a designer may be able to
manipulate bidders’ reference points to his
advantage.
Section 6 considers findings in exploitative
contracting, the study of contract designs
whose central consideration is exploiting an
agent’s mistakes. In most of these models,
the agent is a consumer who either does not
fully understand all features (e.g., all prices
and fees) of a product or mispredicts her own
behavior with respect to the product, and the
principal is a profit-maximizing firm aware of
the consumer’s tendencies. This literature has
reinvigorated the long-recognized but understudied topic of contracting with n
oncommon
priors, but because of its conceptualization
of noncommon priors as resulting from consumer mistakes, its focus on specific forms
of mistakes, and its resulting ability to make
welfare statements, it is usefully categorized
as a new topic in contract theory.
In section 7, I discuss two nascent areas
of research: incomplete contracts and environment design. In the existing research
on incomplete contracts, authors assume
that contracts change preferences by affecting feelings of entitlement or the reference
point to which parties later compare outcomes, and study the implications for optimal contracting and the optimal allocation of
property rights. Finally, section 8 provides a
few general thoughts on the state of the literature and the challenges ahead for it.
1078
Journal of Economic Literature, Vol. LII (December 2014)
2. Key Theories in Psychology
and Economics
This section summarizes the main theories
of individual decision making in psychology
and economics that are used in the applications below. Rabin (1998) and DellaVigna
(2009) provide excellent reviews of the evidence for these models.3 To simultaneously
introduce the model of human behavior, give
a sense of how to work with it, as well as to
illustrate some of its key economic implications, I present each theory together with an
application covered in this review. For brevity, I do not carefully introduce all necessary
assumptions in each example, nor present
formal proofs.
2.1 Loss Aversion
Model. As has been demonstrated in a powerful body of work starting with Kahneman and
Tversky (1979) and Tversky and Kahneman
(1991), individuals evaluate economic outcomes not just according to an absolute valuation attached to the outcomes in question,
but also relative to subjective reference points.
The most important property of such reference-dependent preferences is loss aversion:
losses relative to a reference point are more
painful than equal-sized gains are pleasant.
I present a formal model based on Kőszegi
and Rabin (2006, 2007), assuming here that
the decision concerns only monetary outcomes such as a wage. If the agent’s wage
is w and her reference point is r, then her
reference-dependent utility is
(1) u(w | r)
w + η(w − r) if w ≥ r, and
=
{w + ηλ(w − r) if w < r,
3 Note, however, that the ability of these models to
explain important economic phenomena as explained in
this review can—to the extent that other models cannot
as convincingly account for the same phenomena—be
thought of as additional evidence for the models.
where η > 0, λ > 1. The first term in the
utility function is consumption utility (linear in this example), which corresponds to
the conventional notion of outcome-based
utility. The second term is gain–loss utility,
which captures the agent’s sensations of gain
or loss relative to the reference point. The
parameter λ is the degree of the agent’s loss
aversion—with λ > 1 capturing that losses
are more painful than gains are pleasant—
and η is the weight on gain–loss utility. If
there are other outcomes for which the agent
evaluates gains and losses separately—for
instance, the object in an auction—a similar
additively separable utility function can be
used to capture reference-dependent utility
in those dimensions.
A crucial issue in using loss-averse preferences is the determination of the reference
point r. The most common approach in the
applications below is to set the reference
point equal to the agent’s rational expectations as defined by her full probabilistic
beliefs (Kőszegi and Rabin 2006, 2007). It
is then necessary to extend the above utility
function to allow for the reference point to
be a distribution F(r):
U(w | F) = ∫ u(c | r) dF(r).
(2)
This formulation captures the notion that the
sense of gain or loss from a given consumption outcome derives from comparing it to all
outcomes possible under the reference lottery. For example, if the reference lottery is
a gamble between $0 and $100, an outcome
of $50 feels like a gain relative to $0 and like
a loss relative to $100, and the overall sensation is a mixture of these two feelings.
Application. I illustrate some implications
of the model in the context of wage setting,
beginning with a property of risk attitudes
that reappears in multiple applications.
Suppose w L ≤ w M ≤ w H, and let F denote
the lottery that pays w L with probability
1079
Kőszegi: Behavioral Contract Theory
p L, w M with probability p M, and w H with
probability p H. What is the agent’s expected
utility from F? An agent with rational expectations correctly anticipates the distribution
of outcomes, so her reference point will be
F. Equations (1) and (2) then imply:
U(w L | F) = w L − p M ηλ(w M − w L)
− p H ηλ(w H − w L)
U(w M | F) = w M + p L η(w M − w L)
− p H ηλ(w H − w M)
U(w H | F) = w H + p L η(w H − w L)
+ p M η(w H − w M)
Hence, expected utility is
(3) U(F | F)
= p L U(w L | F) + p M U(w M | F) + p H U(w H | F)
= p L w L + p M w M + p H w H
− η(λ − 1)[ p L p M(w M − w L)
+
p L p H(w H − w L)
+
p M p H(w H − w M)].
8
<0 unless w L = w M = w H
The agent’s expected gain–loss utility is
always negative for risky outcomes, making
her averse to risk. Intuitively, for instance,
when w H > w M, the agent experiences w M
as a loss relative to w H, and w Has a gain relative to w M; but since the sensation of loss
outweighs the sensation of gain, her average
gain–loss sensation is negative. Furthermore,
the agent is first-order averse to risk: her utility decreases linearly with the dispersion in
outcomes, for instance linearly in w H − w M.
In particular, the agent greatly values contracts that insure her from risk on any part of
the distribution of outcomes. These properties of the agent’s aversion to risk are distinct
from the properties generated by classical
expected utility over wealth, leading to many
qualitatively different predictions in economic environments.
As an example of the implications for contracting, I illustrate a loss-aversion-based
explanation for “bonus” contracts—binary
payment schemes in which the employee
has a base salary, and, depending on whether
her performance exceeds a threshold, may
in addition receive a bonus—by Herweg,
Müller, and Weinschenk (2010). Suppose
an agent chooses between low effort (e L)
and high effort (e H), where the cost of effort
e i, i ∈ {L, H}, is c i. The agent’s output takes
one of three values; low, medium, or high.
The following table gives the probabilities
for the three levels of output occurring as a
function of effort:
e L
e H
low
output
medium
output
high
output
2/3
1/3
0
1/3
1/3
1/3
I assume that parameters are such that the
principal would like the agent to exert the
high level of effort. Then, if the agent is
averse to risk, the informativeness principle
in the classical model of moral hazard predicts that the principal pays different wages
for different levels of output (Holmstrom
1979). Intuitively, the three levels of output are differently informative regarding
whether the agent exerted high effort, and
hence should be rewarded differently.
In contrast, the same is not optimal for
a loss-averse agent (even though she is
averse to risk). Let the wage paid for low,
medium, and high output be w L, w M, and
1080
Journal of Economic Literature, Vol. LII (December 2014)
w H, respectively. Clearly, the principal wants
w H ≥ w M ≥ w L. Using equation (3), the
principal’s problem is4
min __
1 (w L + w M + w H)
3
1
__
(PC) s.t. (w L + w M + w H)
3
1 η(λ − 1)(w − w + w
− __
H
M
M
9
− w L + w H − w L)
− c H ≥ u
(IC) __
1 (w L + w M + w H)
3
− __
1 η(λ − 1)(w H − w M + w M
9
− w L + w H − w L)− c H
2 w + __
1 w
≥ __
3 L 3 M
− __
2 η(λ − 1)(w M − w L) − c L
9
The principal minimizes the expected wage
paid to the agent when she exerts high effort,
subject to two constraints. The participation constraint (PC) captures that the agent
must be willing to take the contract over
her best alternative, which has utility u. The
incentive-compatibility constraint (IC) captures that the agent must be willing to exert
high effort over low effort. Rewriting the
constraints:
(PC) __
1 (w L + w M + w H)
3
− __
2 η(λ − 1)(w H − w L) ≥ c H + u
9
4 Technically, in this example there are two dimensions
of utility, money and effort, so one must also account for
gain–loss utility in effort. In this case, however, gain–loss
utility in effort is zero, so that total utility in effort is equal
to consumption utility: when the agent chooses effort i,
with rational expectations both her reference point and her
outcome in the effort–cost dimension are c i.
(IC) __
1 (w H − w L)
3
− __
2 η(λ − 1)(w H − w M) ≥ c H − c L
9
Now I show that w H > w M > w Lis not optimal by identifying a better contract if this is
the case. First, increase w Mby ϵ and decrease
w Hand w Lby ϵ/2. As a result of this, profits
are unchanged, PC is still satisfied, and IC
becomes slack. Then, decrease w Hby ϵ′ and
increase w L by ϵ′. This again leaves profits
unchanged, and for a sufficiently small ϵ′, IC
remains slack, while now PC becomes slack
as well. With both constraints slack, the principal can decrease all three wage levels by
a sufficiently small ϵ″, increasing profits and
still satisfying both constraints.
The above argument implies that the
principal chooses at most two different
wage
levels. Clearly, one wage level (i.e.,
a constant wage) is not incentive compatible, so in the optimal contract the principal
uses exactly two wage levels. Hence, either
w L = w M < w H or w L < w M = w H. Notice
that the above improvement applies when
w L = w M < w H, so this is not optimal either.
Hence, in the optimal contract w L < w M
= w H.
Intuitively, because a loss-averse agent
strongly dislikes random variation in the
wage due to uncertainty in the environment,
the principal has an incentive not to vary
wages too finely with output. At the same
time, he also needs to provide incentives, so
the agent’s wage cannot be completely flat.
As a result, the principal chooses the minimal amount of wage variation that can still
provide incentives: two wage levels, that is, a
bonus contract.
2.2 Present Bias and Time Inconsistency
Model. In intertemporal decisions, many
individuals value the future significantly less
than the present, yet place similar weights on
different points in time in the future. This
Kőszegi: Behavioral Contract Theory
preference implies that a person would not
like to make investments into the future
today, but would like to make similar investments in the future. Time inconsistency—a
conflict between the person’s preferences
at different points in time—arises because
once the future comes, she would again prefer not to make immediate investments.
Laibson (1997) and O’Donoghue and
Rabin (1999a) formalize the above “present bias” using the β − δ model. They
assume that self t (the decisionmaker’s
period-t incarnation) evaluates the stream
u t, u t+1, u t+2, u t+3, … of instantaneous utilities at times t, t + 1, t + 2, t + 3, … as
u t + βδu t+1 + βδ 2 u t+2 + βδ 3 u t+3 + ⋯,
where β, δ ≤ 1 are the discount factors that
parameterize present bias. In this formulation, the discount factor between periods t
and t +1 is βδ, whereas the discount factor
between any two consecutive future periods
is δ. For β < 1, therefore, self t discounts
between the present and the future more
heavily than between different periods in
the future, capturing the above pattern of
intertemporal preferences. For β = 1, the
model reduces to the exponential discounting model commonly used in economics.
A key consideration in applications of the
β − δ model is whether the agent correctly
anticipates the present bias that she will
(but would prefer not to) have in the future.
Following O’Donoghue and Rabin (2001),
I suppose that the agent believes that her
future β will be β̂ . The assumption β = β̂
corresponds to full sophistication regarding
time inconsistency—whereby the agent perfectly understands that she will behave differently than she would now like—whereas
β ̂ = 1 corresponds to full naïveté—whereby
the agent completely ignores her time
inconsistency in preferences. Intermediate
values of β̂ capture intermediate values of
naïvete.
1081
Application. I present a model of credit
contracts based on Heidhues and Kőszegi
(2010) and Eliaz and Spiegler (2006).
Suppose a consumer with β = 1/2 and δ = 1
is contracting in period 0 with competitive
suppliers of credit, who have funds available
to them at an interest rate of zero. The consumer borrows for future consumption and
repays her loan in periods 1 and 2. A credit
contract consists of a borrowed amount c and
a menu of installment plans r 1 ≥ 0, r 2 ≥ 0
according to which the consumer can repay
her loan in periods 1 and 2. Once the consumer signs a contract with a firm, she cannot borrow from other firms, but she can
decide in period 1 which of the installment
plans designated in her contract to follow.
The consumer’s instantaneous utility from
consumption is u(c), whereas her instantaneous disutility from making a payment is
r. Hence, since borrowing is for future consumption, the consumer’s utility in period 0
is u(c) − r 1 − r 2; but in period 1, she chooses
the installment plan to follow minimizing
r 1 + r 2/2.
To find the optimal contract, consider first
fully sophisticated consumers, for whom
β ̂ = β. Since a sophisticated consumer
knows how she will behave, without loss of
generality we can suppose that the menu of
installment plans specifies only the one plan
r 1, r 2 she will choose. Hence, a firm’s problem is
(PC)
max r 1 + r 2 − c
s.t. u(c) − r 1 − r 2 ≥ u,
where u is the consumer’s utility from her
outside option, and the constraint (the “participation constraint” or PC) captures that
the consumer must be willing to accept the
principal’s contract over her best alternative. Substituting the binding PC into the
maximand produces the first-order condition
u′(c) = 1, which means that the c onsumer’s
1082
Journal of Economic Literature, Vol. LII (December 2014)
consumption is efficient. This simple analysis leads to an important point made by
DellaVigna and Malmendier (2004). Even
though the consumer suffers from a timeinconsistency problem, the competitiveequilibrium contract maximizes her utility,
as it leads her to borrow the optimal amount.
Hence, with sophisticated consumers, markets can solve time-inconsistency problems.
The situation is entirely different for a
nonsophisticated borrower, for whom β̂ > β.
To see this, suppose first that c is fixed, so
that the contract concerns only the repayment terms. For any contract the consumer
is offered, there is a “baseline” installment
plan (r ̂ 1 , r ̂ 2 ) she believes in period 0 she will
follow, and a possibly different, “alternative” installment plan (r 1, r 2) she will actually follow. Without loss of generality, we can
assume that the contract contains no other
repayment options. The installment plans
(r 1, r 2) and (r ̂ 1 , r ̂ 2 ) must solve
max r 1 + r 2 − c
(PC) s.t. u(c) − r ̂1 − r ̂2 ≥ u
(PCC)
(IC)
r ̂1 + β̂ r ̂2 ≤ r 1 + β̂ r 2
r 1 + r 2/2 ≤ r ̂1 + r ̂2/2
As explained by Heidhues and Kőszegi
(2010), the firm faces the following constraints in designing its contract. First, for
the borrower to be willing to accept the
firm’s offer, self 0’s utility given how she
believes she will behave must be at least u
(PC). Second, if self 0 is to think that she will
choose the baseline option, then given her
beliefs β̂ she must think she will prefer it to
the alternative option (perceived-choice constraint or PCC). Third, if self 1 is to actually
choose the alternative repayment schedule,
she has to prefer it to the baseline repayment
schedule (IC).
The solution to this problem is (r ̂1, r ̂2)
= (u − u(c), 0) and (r 1, r 2) = (0, 2(u − u(c))).5
Given that in a competitive equilibrium
firms make zero profits, the equilibrium
payments are (r ̂1, r ̂2) = (c/2, 0) and (r 1, r 2)
= (0, c). In practice, this type of contract
would correspond to a credit arrangement in
which, as for instance with many credit cards
and n
ontraditional mortgages, the consumer
is asked to repay her loan fast, and delaying
repayment carries large penalties.
The optimal contract, therefore, induces
false beliefs in the consumer regarding how
she will repay, making the contract look more
attractive than it really is. In fact, the contract
maximizes the consumer’s mistake: by asking
her to repay only in period 1, the lender maximizes the effect of the consumer’s mistaken
beliefs regarding her future willingness to
pay to delay repayment. This first implication of the example, whereby consumers are
led to buy seemingly cheap products, is a
general implication of exploitative contracting that appears in many papers discussed in
section 6.
The fact that consumers are sold seemingly cheap products can have many types
of adverse welfare consequences. In this
particular case, a welfare cost arises if the
borrowed amount c is endogenous. Since
the consumer believes that borrowing c costs
only c/2, the competitive equilibrium with
endogenous c has u′(c) = 1/2—the consumer overborrows relative to the optimum.
Note that since the borrowing is for future
consumption, the overborrowing is not due
to the borrower’s time inconsistency per se,
but due to her false beliefs about her future
repayment behavior.
5 For any (r ̂ , r ̂ ), (r , r ) = (0, 2 r ̂ , r ̂ ) satisfies both IC
1 2
1 2
1 2
and PCC (the latter constraint is slack), and no other (r 1, r 2)
that satisfies IC achieves higher profits, so it is the optimal
(r 1, r 2). Then, among (r ̂1, r ̂2) that satisfy PC, the option
that maximizes the principal’s resulting profits, 2 r ̂1, r ̂2,
has r ̂2 = 0.
Kőszegi: Behavioral Contract Theory
All of the above holds for any β̂ > β—
that is, for even arbitrarily small degrees of
naïveté regarding time inconsistency. This
leads to the second important implication
of the example: with time inconsistency,
even small amounts of naïveté often have
large welfare implications. Heidhues and
Kőszegi (2010) show that simple restrictions
on the contract form can drastically increase
welfare. As an illustration, suppose that we
impose a sufficiently low constraint p on how
much the consumer can be penalized for failing to choose the lowest-cost installment plan
in her contract. Then, in one competitiveequilibrium contract for a fixed consumption level c, the consumer chooses between
installment plans (c − p, 0) and (0, c).6 Since
the consumer thinks the cost of borrowing c
is c − p, the competitive-equilibrium level of
borrowing becomes u′(c) = 1. Hence, with
this regulation the consumer borrows the
optimal amount, and—although she falsely
believes she will get a better deal—ultimately repays exactly that amount.
2.3 Inequity Aversion
Model. To explain a number of disparate
facts regarding prosocial behavior, Fehr and
Schmidt (1999) present a model in which
individuals dislike both advantageous and
disadvantageous inequality in material outcomes, but dislike being behind more than
they dislike being ahead. Suppose that there
are N individuals in agent i’s “reference
group”—the group of individuals (including
herself) whose material outcomes she cares
about. Then, her utility from material payoffs x 1, … , x Nis
(4)
x i − ______
1 α i ∑
( x − x i) + β i ∑
(x i − x j) ,
N − 1 ( j;x j≥x i j
)
j;x j<x i
6 In this case, there are other competitive-equilibrium
contracts: for instance, if r ̂1 ≥ p and r ̂1 + r ̂2 = c − p, then
(r ̂1, r ̂2) and (r 1, r 2) = (r ̂1 − p, r ̂2 + 2p) is a competitive
equilibrium.
1083
where β i satisfying 0 ≤ β i < 1 parameterizes the agent’s aversion to being ahead and
α i ≥ 0 parameterizes her aversion to being
behind.
Application. I illustrate the implications of
inequity aversion for incentive contracts.
First, consider a “gift-exchange” situation,
in which the principal pays the agent a fixed
output-independent wage w ≥ 0, and then
the agent chooses effort level e ≥ 0. The
agent’s material cost of effort is c(e), and this
produces output e for the principal. Hence,
the agent’s material payoff is x 2 = w − c(e),
whereas the principal’s is x 1 = e − w, and
the effort level that maximizes social welfare
satisfies c′(e) = 1. Note that if the principal
and the agent were selfish, the agent would
exert zero effort for any wage, and hence the
principal would pay a wage of zero.
As has long been recognized in the literature, however, inequity aversion can lead to
a higher wage and a higher level of effort,
potentially increasing material welfare for
both the principal and the agent. Suppose
that the principal and the agent are inequity
averse as defined above, N = 2, the principal is player 1 and the agent is player 2, and
α i ≥ β i > 1/2. We first solve for how the
agent reacts to a wage w > 0, assuming that
the optimal e has c′(e) ≤ 1 (the analysis below
makes clear that this is the case for the principal’s optimal contract). Clearly, the agent
never chooses an effort level that leads to
x 1 > x 2; relative to x 1 = x 2, this would lead
to a lower material payoff for her and generate further disutility from disadvantageous
inequality. Hence, the agent maximizes
x 2 − β 2(x 2 − x 1) = w(1 − 2β 2) + β 2 e −
(1 − β 2)c(e) subject to the constraint x 2 ≥ x 1,
or w − c(e) ≥ e − w. Now it is easy to see
that w − c(e) > e − w cannot be optimal: if
this were the case, then (using that β 2 > 1/2
and therefore β 2 > 1 − β 2, and c′(e) ≤ 1)
the agent would want to increase e. As a
result, the agent’s inequity aversion leads
1084
Journal of Economic Literature, Vol. LII (December 2014)
her to choose her level of effort to equalize
material payoffs, setting w − c(e) = e − w,
or w = (e + c(e))/2.
Given that the agent’s behavior eliminates
inequality, the principal maximizes material
payoffs, and her problem can therefore be
written as
(5)max e − w
e + c(e)
.
s.t. w = _______
2
Substituting the constraint into the maximand yields that the principal is maximizing
(e − c(e))/2, leading to c′(e) = 1. This means
that inequity aversion generates an efficient
outcome: it induces an effort level that maximizes material efficiency, and since this goes
along with the lack of inequality, it is also
efficient in terms of inequity-averse preferences. To understand the intuition, note that
the agent’s inequity aversion acts as an incentive device: since a higher wage means that a
higher effort is necessary to split the surplus
equally, the agent responds to a higher wage
by increasing effort. Knowing this and knowing that the agent will give him exactly half of
the social surplus, the principal has an incentive to induce the effort level that maximizes
social surplus.
An alternative to gift exchange is the “voluntary-bonus” system, in which the agent first
chooses her level of effort, and then the principal chooses a bonus b ≥ 0 to reward her.
Similarly to how the agent chooses her level
of effort in the case of gift exchange above,
the principal chooses the bonus to equalize
material payoffs, setting b − c(e) = e − b, or
b = (e + c(e))/1. Again, therefore, inequity
aversion creates an incentive device, but it
does so in a different way: while the giftexchange arrangement relies on the inequity
aversion of the agent to reward the principal’s investment of a high wage, the voluntary-bonus system relies on the principal’s
inequity aversion to reward the agent’s
investment of a high effort.
As
the
principal’s
bonus-setting
behavior results in no inequality, the
agent
maximizes
material
payoffs
b − c(e) = (e + c(e))/2 − c(e), which again
yields c′(e) = 1. Hence, the voluntary-bonus
arrangement also maximizes social surplus.
The intuition is simple: knowing that the
principal will split the social surplus e − c(e)
equally, the agent has an incentive to maximize social surplus.
The above analysis, however, provides
a simplistic view of the power of inequity
aversion to generate efficient outcomes.
Motivated by Fehr et al. (2007), suppose
that a proportion q of individuals in society are inequity averse as above, but a proportion 1 − q are selfish. In that case, the
gift-exchange and bonus arrangements are
neither optimal nor equivalent. To illustrate,
I suppose for simplicity that α i = ∞, so that
inequity-averse individuals do not tolerate
disadvantageous inequality.
Consider first a gift-exchange arrangement. An inequity-averse agent responds
to a positive wage as above, but a selfish
agent chooses zero effort. Since a positive
wage followed by zero effort generates disadvantageous inequality for the principal,
an inequity-averse principal—afraid to run
into a selfish agent—never chooses a positive
wage. Denoting an inequity-averse agent’s
effort as a function of the wage by e(w),
the selfish principal chooses w to maximize
qe(w) − w, yielding the first-order condition
qe′(w) = 1. Totally differentiating equation
(5) with respect to w, solving for e′(w), and
plugging into the first-order condition gives
c′(e) = 2q − 1.
Hence, effort is lower than optimal for any
q < 1.
Now I return to the voluntary-bonus
arrangement. An inequity-averse agent
Kőszegi: Behavioral Contract Theory
would find it intolerable to exert positive
effort and then not be rewarded by a selfish
principal, so she exerts zero effort. Recalling
that an inequity-averse principal chooses
b = (e + c(e))/2, a selfish agent maximizes
q(e + c(e))/2 − c(e), which gives the firstorder condition
q
c′(e) = _____
.
2 − q
Again, effort is lower than optimal for any
q < 1. The intuition for why effort is lower
than optimal in the presence of selfish individuals under either arrangement is simple:
since a party’s attempt to increase the surplus might go unrewarded, she has less of an
incentive to increase the social surplus.
But while both arrangements generate
lower-than-optimal effort, they do not generate the same effort. Simple arithmetic
shows that q/(2 − q) > 2q − 1, so effort is
higher under the bonus arrangement than
under gift exchange. Furthermore, while
the bonus system generates effort whenever
the agent is selfish, the gift-exchange system
generates effort only if the principal is selfish and the agent is inequity-averse. Hence,
under heterogeneity in inequity aversion,
the bonus system is superior. This difference
is due to a difference in how much the first
mover has to invest in an attempt to increase
the social surplus and be rewarded for it by
an inequity-averse second mover. For the
agent to increase the social surplus under
the bonus system, she must pay her marginal cost of effort. But for the principal to
increase the social surplus by increasing the
wage in the gift-exchange system, he must
compensate the agent for her marginal cost
of effort plus half of the marginal increase in
the social surplus (since when splitting the
social surplus, the agent will keep half of the
increase for herself). When this investment
is certain to be rewarded—when there are
no selfish individuals around—it does not
matter which investment is more costly, as
1085
in either case the investing party gets half
of the increase in the social surplus. When
there is a chance that the investment will not
be rewarded, however, investment is lower
when its marginal cost is higher.
2.4 Quasi-Bayesian Models
Framework. The literature has documented a number of systematic mistakes
individuals commit when thinking about
statistical properties of random phenomena. A commonly used reduced-form way to
capture such mistakes is the quasi-Bayesian
approach: one posits that the agent updates
her beliefs fundamentally in a Bayesian way,
but commits a particular error that is inconsistent with rational inference. These models
typically also assume that the agent does not
learn about her error from her observations.
The quasi-Bayesian approach is not a theory,
but merely a framework for thinking about
statistical mistakes. A fully-fledged theory
specifies the particular mistake the agent
is committing. In the models reviewed in
this article, there are two general mistakes
authors assume: systematically incorrect priors, and mistakes in updating beliefs based
on information. The most common version
of an incorrect prior is overconfidence: the
agent has overly positive views about herself
or some of her prospects.
Application. I consider the implications of
overconfidence for contracting. First, suppose that an agent produces one of two levels
of output, high or low. The true probability
of high output is q, but the agent has an
incorrect prior in that she believes the probability of high output is q̃ > q. The principal offers an output-contingent wage, paying
w H for high output and w L for low output.
The principal is risk neutral, while the agent
has mean-variance preferences of the form
q̃ w H + (1 − q̃ )w L − q̃ (1 − q̃ )(w H − w L)2.
For now, suppose that there is no asymmetric information or moral hazard.
1086
Journal of Economic Literature, Vol. LII (December 2014)
In this case, the principal’s problem is
(PC) min qw H + (1 − q)w L
s.t. q̃ w
H + (1 − q̃ ) w L
− q̃ ( 1 − q̃ ) (w H − w L)2 ≥ u.
H +
Noting that qw H + (1 − q)w L = q̃ w
(1 − q̃ ) w L − (q̃ − q)(w H − w L), we can plug
PC into the maximand to get
min u + q̃ ( 1 − q̃ ) (w H − w L)2
− (q̃ − q)(w H − w L).
This gives the optimal wage spread
(6)
̃ − q
q
≡ Δ*
w H − w L = _________
2q̃ ( 1 − q̃ )
Given w H − w L, the principal sets the levels
of the two wages so that PC binds.
Equation (6) identifies a basic exploitation
motive with overconfident agents: if the agent
is too optimistic that high output will occur,
she overvalues the wage paid for high output, so the principal can decrease expected
wages by paying her a little more for high
output and much less for low output. This
is a special case of the well-known observation that individuals with different priors
would like to speculate against each other
(Harrison and Kreps 1978, for instance). It is
also an example of a situation where a type of
contract that may appear to have a classical
purpose (an output-contingent wage often
serves to overcome moral hazard) instead
has an exploitative purpose.
The exploitation motive, however, is limited for agents with only slightly overoptimistic beliefs, so that in more realistic settings
these agents do not receive exploitative
contracts. This finding contrasts with that
for time-inconsistent agents above—where
even small degrees of naïveté can have large
welfare effects—suggesting that the consequences of false beliefs can be less serious
under time consistency than under time
inconsistency. As a first example of how
agents with small overconfidence can fare
well, I consider a version of de la Rosa’s
(2011) moral-hazard model. Suppose that the
agent chooses between two levels of effort,
high and low. Her output with high effort is
as described above, but her output with low
effort is low with probability 1. The cost of
low effort is zero, and the cost of high effort
is e. The agent’s incentive-compatibility constraint is then
q̃ w
H + (1 − q̃ ) w L
− q̃ ( 1 − q̃ ) (w H − w L)2 − e ≥ w L.
or
(7) q̃ ( w H − w L) − q̃ ( 1 − q̃ ) (w H − w L)2 ≥ e.
Let the lowest w H − w L satisfying this
inequality be Δ**. The comparison of Δ*
and Δ** highlights an important distinction
regarding different kinds of contracts in this
review. If Δ* > Δ**, then w H − w L = Δ*
is optimal, and the agent’s IC constraint
is not binding. In the sense that the principal’s primary consideration is exploiting
the agent’s mistaken beliefs—and possible
constraints derive from this attempt, while
other constraints are either nonexistent or
not binding—the contract is exploitative
by the (somewhat informal) standards of
Section 6. In contrast, if Δ* < Δ**, then
w H − w L = Δ** is optimal. In this case,
therefore, the consideration determining
w H − w L remains to make sure that the
agent’s incentive-compatibility constraint
is satisfied, not to take bets against her. In
this sense, the contract is not primarily about
exploitation, although the agent’s expected
wage is still affected by her overconfidence.
Kőszegi: Behavioral Contract Theory
It is interesting to note that—as is easy to
check using equation (7)—in this latter case
the optimal wage spread is decreasing in
overconfidence. Intuitively, an overconfident
agent overestimates how easily she can produce high output, and hence a less outputsensitive incentive contract is sufficient to
induce her to choose high effort. This insight
has some interesting welfare implications.
First, because a lower-powered incentive
contract allows the agent to bear less risk,
mild overconfidence increases efficiency.
Second, when principals compete for the
agent, the agent enjoys the full increase in
social efficiency, so that having slightly wrong
beliefs actually benefits her.7
As a second limitation on exploiting small
degrees of overconfidence, consider screening agents with different degrees of optimism, based on Eliaz and Spiegler (2008).
Suppose that there is no moral hazard, but
in addition to the agent above, there is an
agent with beliefs Q̃ > q̃ , and the principal
would like to contract with both agents without directly observing their beliefs. Based on
equation (6), the principal would like to offer
a slightly exploitative contract to the less optimistic agent q̃ and a more exploitative contract to the more optimistic agent Q̃ . Given
the latter incentive, however, the principal
might not exploit agent q̃ . I demonstrate
this heuristically by showing that the principal prefers to offer agent q̃ a contract with
w H − w L = 0 rather than w H − w L = ϵ,
where ϵ > 0 is small. With w H − w L = ϵ,
the less optimistic agent’s binding PC
becomes
w L + q̃ ϵ − q̃ ( 1 − q̃ ) ϵ2 = u,
7 To see this, note that under competition, principals
pay the same expected wage to any agent—the agent’s
expected output conditional on high effort—so a decrease
in the wage spread increases the agent’s utility.
1087
which for small ϵ implies w L ≈ u − q̃ ϵ ,
yielding qw H + (1 − q)w L = w L + qϵ ≈ u −
(q̃ − q)ϵ. Hence, relative to a contract with
w H − w L = 0, the principal makes a profit of
approximately (q̃ − q)ϵ on agent q̃ . But how
much does she lose on agent Q̃ ? To see this,
note that agent Q̃ ’ s perceived value from taking the above contract is
w L + Q̃ ϵ − Q̃ (1 − Q̃ )ϵ2
≈ w L + Q̃ ϵ ≈ u + (Q̃ − q̃ ) ϵ
This means that, when contracting with agent
Q ̃ , the principal loses a profit of (Q̃ − q̃ ) ϵ
because having the less speculative contract around improves agent Q̃ ’ s perceived
alternative option. If q̃ − q is sufficiently
lower than Q̃ − q, therefore, exploiting agent
q ̃ is not worth it for the principal. Intuitively,
while the less optimistic type values a contract with w H > w Land therefore allows the
principal to decrease expected wages, the
more optimistic type values the same contract much more, so that—to avoid having to
pay a kind of information rent—the principal
would rather not exploit the less optimistic
type. This result implies that the principal
either does not exploit the agent or exploits
her substantially, generating the empirical implication that if we see exploitation, it
should be large.
3. Moral Hazard
For the rest of this review, I turn to a
more comprehensive, but informal, survey
of efforts to incorporate psychology-andeconomics ideas into contract theory. I organize the topics according to classical contract-theoretic principles, and begin with
discussing moral hazard.
Moral hazard arises in contracting situations when a decision taken by the agent
after the contract is signed affects the utility
1088
Journal of Economic Literature, Vol. LII (December 2014)
of the principal, and the action cannot be
directly contracted upon. In these settings,
there are typically outcomes related to the
agent’s action that can be contracted upon
(e.g., output), and the key issue is how
the principal optimally designs a contract
that induces the agent to take an action he
prefers (captured in the incentive-compatibility constraint), and that the agent
will accept (captured in the participation
constraint). This section summarizes the
existing behavioral-economics literature on
moral hazard.8
3.1 Intrinsic Motivation: Reliance on
Voluntary Actions
One important theme that has emerged
from the literature is that optimal incentive schemes may rely crucially on “intrinsic motivation”—actions that do not carry a
financial reward. This contrasts with classical models of moral hazard, where the only
source of motivation is explicit or implicit
financial incentives.
Social preferences as a source of intrinsic
motivation. Akerlof (1982) was perhaps the
first (within economics) to argue that firms
might pay high wages to induce high effort
as part of a gift exchange. Englmaier and
Leider (2012) formalize this idea in a model
in which the agent has inequity-averse
preferences vis-à-vis the principal. As in
previous models, the principal can induce
high effort with performance-based pay.
But he might also induce high effort with
performance-independent pay. Specifically,
if the principal pays a low (near-marketclearing) fixed wage, the agent does not
work hard, lest the principal enjoy a much
8 Kamenica (2012) reviews evidence of many systematic
psychological forces on how individuals respond to incentives, and more specifically, Fehr, Goette, and Zehnder
(2009) review evidence on how reference-dependent fairness concerns affect interactions in a labor-market setting.
higher payoff; but if the principal pays a
high fixed wage, the agent responds with
high effort so that the principal receives a
fair share of the pie. Hence, a high wage is
often profit maximizing because it puts the
agent ahead and thereby creates intrinsic
motivation to help the principal. Englmaier
and Leider (2012) also investigate conditions under which it is optimal to use
inequity-aversion-based incentives versus
performance-based pay, and show that
inequity aversion is more likely to be used
if output is a poor signal of effort or the
agent is highly inequity averse or productive. This is one formalization of Akerlof’s
distinction between “primary” labor markets (governed by gift exchange and abovemarket-clearing wages) and “secondary”
labor markets (governed by explicit incentives and market-clearing wages).
Voluntary
bonuses.
Fehr,
Klein,
and Schmidt (2007) establish that if
heterogeneity in inequity aversion is
present in the population, a contract
based on voluntary bonuses—by which the
principal can reward the agent for high
effort ex post—often dominates the above
high-wage contract based on gift exchange.
A bonus contract engages inequity aversion
much like a high-wage contract above,
except with the order of moves reversed:
much like an i
nequity-averse worker
responds to a high wage (and only a high
wage) with high effort, an inequity-averse
principal is willing to pay the bonus in
response to (and only in response to) high
effort. Hence, if there are sufficiently many
fair-minded principals, both fair-minded
and selfish agents are willing to exert effort.
The superiority of the bonus scheme arises
from the presence of selfish individuals
in the population: a party that first takes
a costly action (exerting effort or paying a
high wage, respectively) faces the risk that
a selfish partner will not reward it, and this
Kőszegi: Behavioral Contract Theory
risk is lower for the bonus contract because
the action has lower cost in that case.9
3.2 Reduced Wage Sensitivity
Another finding in the literature is that—
for multiple reasons—an agent’s wage may
be less responsive to her output than a classical moral-hazard model predicts.
Wage compression. Complementing the
research on moral hazard with social preferences reviewed in the previous subsection,
it seems natural to assume that employees compare themselves not just to their
employer, but also to fellow employees. A
very intuitive implication of inequity aversion is then that wages tend to be more compressed than what one would expect based
on classical models. Because agents do not
like the ex post inequality that can occur by
chance as a result of high-powered (i.e., very
performance-sensitive) individual incentives, to relax their participation constraint
individual incentive pay is reduced. When
agents are sufficiently averse to ex post
inequality, the principal might institute
team-based incentives not because he cannot observe individual performance or wants
to encourage cooperation (as in many classical models), but to insure agents against
painful inequality (Englmaier and Wambach
2010; Bartling 2011). Furthermore, Bartling
and von Siemens (2010) argue that because
comparisons are less pronounced across
firms than within firms, consistent with the
evidence inequity aversion predicts no wage
compression across firms.
9 Similarly, Fehr and Schmidt (2004) show that a
voluntary-bonus scheme outperforms an explicit incentive contract in a multitasking environment in which performance on both tasks is observable to the principal, but
performance on only one is contractible. As is well-known
since Holmström and Milgrom (1991), providing incentives on the contractible task leads an agent to ignore the
unincentivized task, leading to inefficiency. A scheme in
which the principal can reward high effort on both tasks
with a voluntary bonus increases efficiency.
1089
Despite the general tendency of inequity
aversion to generate wage compression, Itoh
(2004), Rey-Biel (2008), and Bartling and
von Siemens (2010) identify some conditions under which increased pay inequality
can result. Specifically, if limited liability prevents the principal from punishing the agent
by reducing her pay, he can instead punish
her by paying another agent more. In ReyBiel’s (2008) two-agent deterministic setting, for instance, an agent who works hard
when the other agent shirks is more than
compensated for her cost of effort. In other
words, the principal may prefer tournamenttype incentives even when classical reasons
to prefer such an incentive structure—e.g.,
common shocks—are absent.
Simple schemes due to loss aversion. In the
canonical model of moral hazard in a risky
environment, increasing the power of incentives is beneficial to the principal because
it aligns the agent’s motives with the principal’s goals, but it is also costly because it
requires exposing the agent to more risk and
hence paying her a higher risk premium.
Because loss aversion affects a person’s attitudes toward risk, it affects this fundamental
tradeoff. A finding that reappears in multiple papers is that—because it generates
first-order risk aversion—loss aversion leads
to the unresponsiveness of transfers to outcomes over some regions of outcomes. As
an extreme example in the context of contracting when the consumer is loss averse
and her demand for the product is uncertain
(e.g., mobile phone contracts), Herweg and
Mierendorff (2013) show that—consistent
with many real-world examples—the seller
might prefer a flat-rate contract. The flat-rate
contract leads the consumer to overuse the
product, but it is still profit maximizing if this
moral-hazard problem is sufficiently weak.
In many moral-hazard settings, especially
employment contracts, however, incentives
must also be provided, so completely flat
1090
Journal of Economic Literature, Vol. LII (December 2014)
contracts are not optimal. In such situations,
wages must be responsive to performance
somewhere, but nevertheless, loss aversion
predicts simpler reward schemes than do
typical classical settings. Herweg, Müller,
and Weinschenk (2010) establish that if the
agent’s reference point is her rational expectations about the wage (as in Kőszegi and
Rabin 2006, 2007), then the optimal contract
often has a “bonus structure,” with two possible wage levels. In a dynamic extension of
the expectations-based model due to Kőszegi
and Rabin (2009), Macera (2012) proves that
under some circumstances the current wage
is completely unresponsive to performance,
with incentives provided by future opportunities.10 And de Meza and Webb (2007) show
that if the reference point is the median
wage, then the agent’s wage is unresponsive
to performance up to the median performance level, but responsive to performance
above that level.11
Wage stickiness. Eliaz and Spiegler (2014)
develop a model of labor-market dynamics that combines the two major features
of worker behavior studied separately by
the papers above: (i) that workers can be
induced to exert discretionary effort by paying them a fair wage; and (ii) that workers’
notion of a fair wage is based on their lagged
10 In Daido and Murooka (2012), team incentives
emerge under some circumstances as an optimal way to
manage agents’ loss aversion, even when individual incentive pay is feasible. In particular, suppose that a hard-working agent’s probability of success in her project is low, and
consider first individual incentive pay that is based on the
performance of only her own project. Then, working hard
creates an expectation of high pay and thereby exposes the
agent to a likely sense of loss, dampening the incentive to
work. By paying the agent also for other agents’ successes,
the principal reduces the probability of loss, increasing the
incentive to work. From a single agent’s point of view, getting paid for another agent’s success is equivalent to the
principal sometimes “turning a blind eye”—i.e., forgiving
failure—but Daido and Murooka argue that team incentives are a credible way to implement such a blind-eye
policy.
11 For related work, see also Jofre et al. (2012).
expectations of the wage. In each period,
firms face a productivity shock and make a
one-period fixed-wage take-it-or-leave offer
to workers. The subgame-perfect equilibrium displays a number of interesting and
economically important features. First,
because firms are reluctant to cut the wage
below the reference wage, workers already
inside the firm experience downward wage
rigidity: in response to moderate decreases in
productivity, the wage remains unchanged.
Second, if productivity falls too much to
pay the reference wage, firms do not induce
discretionary effort. Then, depending on
the importance of discretionary effort, they
either retain the worker with a low wage
knowing that this will result in low effort, or
lay the worker off, both of which are inefficient. Third, since newly hired workers had
expected to be employed with lower probability, they come in with lower expectations
regarding how much they will make, so their
wage is more flexible than that of existing
workers.
3.3 The Effect of Extrinsic Motivation on
Intrinsic Motivation
As mentioned in section 3.1, a number
of theories in behavioral economics imply
that an agent might be willing to exert costly
effort for reasons other than explicit material incentives. At the same time, all theories
assume that agents are also responsive to
explicit incentives. The recognition that an
agent may be motivated by different types of
incentives raises the natural question of how
different incentives interact. A number of
papers in the literature suggest that extrinsic incentives can crowd out intrinsic motivation, so that it can be optimal to employ
weaker explicit incentives than classical
models would suggest.
In addition to the stylized fact that explicit
performance incentives are often quite
weak, different strands of research in this
area are motivated by (and consistent with)
Kőszegi: Behavioral Contract Theory
somewhat different kinds of empirical observations. Some researchers have found circumstances under which explicit incentives
have a contemporaneous negative effect on
effort—that is, effort is lower when greater
explicit incentives are in place. In particular, Gneezy and Rustichini (2000) document
that imposing a modest fine on parents for
picking up their kids late at a day-care center
increases the prevalence of late pickups. And
Falk and Kosfeld (2006) find in an experiment that a “controlling contract”—which
prohibits the lowest contribution levels
by the agent—also lowers contributions.12
This evidence is slightly different from that
studied extensively since Deci (1975) and
reviewed for instance by Deci, Koestner, and
Ryan (1999), whereby providing extrinsic
incentives to perform a task leads individuals to engage in the same or similar tasks less
often once the incentives are removed.
Given the topic of this subsection, I organize existing research according to the mechanism through which explicit incentives act
on intrinsic motivation.
Informed principals. A number of theories
employ an informed-principal framework in
which the principal has superior information regarding a variable that affects both
his incentive-design problem and the agent’s
willingness to exert effort, so that a contract
with strong explicit incentives can undermine
the agent’s motivation through its effect on
her beliefs. In Bénabou and Tirole (2003),
the principal chooses a bonus for task completion knowing the agent’s ability or cost of
completing the task, while the agent receives
only a signal of this v ariable. If the p
rincipal
12 For the above negative effect to occur, it seems crucial that the control is imposed by the principal rather than
exogenously. For instance, List (2007) finds that when allocating money between themselves and a recipient, many
fewer dictators give a positive amount if they have the
option of taking away money than if they do not have this
option.
1091
knows that the cost is high, he is worried
that the agent might also have received a signal that the cost is high, so that he chooses
a higher bonus to motivate the agent. As a
result, a large bonus for completing a task
becomes a signal of task difficulty for the
agent, which reduces her motivation to work
in the future. In a closely related paper with
multiple agents, Fang and Moscarini (2005)
suppose that the principal receives a private
signal about each agent’s ability, and can
decide whether to make different contract
offers for agents with different signals. If
agents are overconfident, the principal might
prefer not to differentiate contracts, as the
information from differentiated contracts
would (given their overconfident prior) provide information to agents that is on average
bad news and, hence, would lower intrinsic
motivation.
Other papers have a similar theoretical
structure to that of Bénabou and Tirole,
but the information the principal has is of
a different nature. In Sliwka (2007), there
are three types of agents in the population:
selfish, fair, and conformist. Selfish agents
care only about their material well-being,
whereas fair agents care also about the principal’s payoff. Conformists behave like the
other type (selfish or fair) that they believe
is in the majority in the rest of the population. The principal knows the share of fair
types, which affects both his optimal incentive scheme and conformists’ willingness to
exert effort. And in Herold (2010), the agent
may be “trustworthy”—motivated to exert
effort even without explicit incentives—or
“untrustworthy”—willing to exert effort only
if explicitly motivated to do so. The principal
has privately held beliefs about the agent’s
probability of being trustworthy (his “trust”
in the agent). This always affects the profitability of different incentive schemes. It can
also affect the agent if she takes nontrusting
behavior as a hostile act and reciprocates it,
or if the principal also takes an action that is
1092
Journal of Economic Literature, Vol. LII (December 2014)
either a complement or a substitute to the
agent’s action.
The above informed-principal models can
naturally explain evidence on the future negative effect of explicit incentives mentioned
at the beginning of the subsection: once the
agent draws the respective negative conclusion from a high-powered incentive contract
in each case, she will be less motivated in
future tasks. But these models do not provide full explanations for the contemporaneous negative effect of explicit incentives.
While the models explain why explicit incentives might be weakened by the crowd-out
of intrinsic motivation, they remain positive
reinforcers in the sense that, if a given worker
receives strong rather than weak incentives
(among equilibrium contract offers), she
exerts greater effort on average. The intuition derives simply from the negative future
effect of explicit incentives: since strong
explicit incentives have negative future consequences, the principal uses them only if
they increase current effort.13
Social signaling. Bénabou and Tirole
(2006) develop a model in which agents have
heterogeneous degrees of altruism, and all
types prefer to be perceived by others as
altruistic. If there are no explicit incentives,
prosocial behavior signals a high degree of
altruism, so that many agents are willing to
act prosocially for image reasons. But with
explicit incentives for prosocial behavior,
such behavior could be due to materialism and is thus a less convincing signal of
altruism, reducing the image motivation
13 Nevertheless, Bénabou and Tirole (2003) point out
that the unconditional correlation between explicit incentives and effort could be negative, which could lead a naïve
observer to incorrectly conclude that extrinsic incentives
crowd out contemporaneous intrinsic motivation. This can
happen because explicit incentives are more likely to be
used when the principal’s information points to low agent
motivation, which is also when the agent is less likely to
work.
for
prosocial behavior. This crowding out
reduces the effectiveness of extrinsic incentives, and under some conditions can induce
a negative net effect of extrinsic incentives.
Bénabou and Tirole’s model explains the
evidence on the contemporaneous negative effect of explicit incentives on prosocial
behavior mentioned above. Since removing
the explicit incentives reinstates the signaling motive, however, the model does not
seem to predict a future effect.
Reciprocal behavior. Another potential
explanation for the positive contemporaneous
impact of weak explicit incentives on effort
is reciprocal behavior. The original model of
intentions-based reciprocity, Rabin (1993),
defines a kind act as one that sacrifices the
person’s own material well-being to increase
the material well-being of the other player,
and assumes that a reciprocal player prefers
to respond to a kind act with a kind act. While
in this model weak incentives cannot generate
an increase in effort if the agent is known to
be reciprocal—if they did, they would benefit
the principal and hence could not be considered kind—von Siemens (2011b) shows that
they can do so under some circumstances if
both selfish and reciprocal agents are present. Since selfish agents take advantage of
weak explicit incentives at the expense of the
principal, choosing such incentives is kind
vis a vis selfish agents. If there are many selfish agents, therefore, choosing weak explicit
incentives is on average a kind act, to which
reciprocal agents respond with high effort.
Furthermore, this is profitable for the principal if reciprocal agents respond sufficiently
strongly to the principal’s kindness. Ellingsen
and Johannesson (2008) demonstrate a similar result in a closely related model that differs in why exactly the agent responds to a
kind act kindly: the principal’s choice of
incentive contract reveals how nice the principal is, and some agents prefer to be nice to
nice principals.
Kőszegi: Behavioral Contract Theory
Anticipatory utility. Immordino et al.
(2011) consider the problem of a risk-neutral principal offering an output-contingent
wage to a risk-neutral agent with unlimited
liability, who receives a private signal about
her return to effort before choosing her
effort level. The agent derives utility from
anticipating high wages, so that she has an
incentive to suppress signals that indicate a
low return to effort. If anticipatory utility is
sufficiently important relative to actual outcomes, then (in contrast to the prediction
of a classical model) implementing the firstbest level of effort is impossible. Intuitively,
the agent exerts high effort only if she is sufficiently rewarded for a good result, but this
makes a good result very desirable, increasing the motive to suppress bad news and
hence distorting information use.
3.4 Moral Hazard and Overconfidence
Individuals often overestimate their ability to do well in a task, either in the sense
of baseline optimism—overestimating one’s
performance given an effort level—or in the
sense of control optimism—overestimating
the return to effort—or both.14 Researchers
have explored a few implications of these
phenomena for moral hazard.
De la Rosa (2011) and Gervais, Heaton, and
Odean (2011) study moral-hazard models in
which the risk-averse agent is both baseline
optimistic and control optimistic. Since a
control-optimistic agent requires lowerpowered incentives than a rational agent to
implement a given level of effort, the principal responds to moderate overconfidence by
lowering the power of incentives. This allows
the agent to bear less risk, increasing social
welfare. Interestingly, if there is competition between principals, the agent receives
all of this increase in social welfare, so that
14 The above distinction is taken from Spinnewijn
(2012).
1093
her bias benefits her in equilibrium.15 In contrast, because baseline optimism implies that
the agent is overly willing to accept highpowered contracts with a low base wage, the
principal responds to large overconfidence
with very high-powered contracts to exploit
the agent’s mistake, and in equilibrium the
agent’s incentive constraint may not bind.
In this range, the contract is more like an
exploitative contract as defined in section 6,
and the large overconfidence always hurts
the agent.16
In the context of unemployment insurance
and job-search behavior, Spinnewijn (2012)
assumes that agents are baseline optimistic
and may be control pessimistic or control
optimistic, and contrasts the response of a
social planner (who maximizes agents’ utility
subject to a budget constraint) and a competitive profit-maximizing firm. When choosing the level of benefits for a biased agent,
the social planner accounts for the effect of
a change in search behavior on the agent’s
welfare, so that he responds to control optimism by increasing benefits to lower search
effort. A profit-maximizing firm, in contrast,
does not care about the agent’s welfare, so
it does not face the same consideration. But
because the agent’s baseline optimism lowers
her demand for insurance, the firm responds
by lowering benefits.
15 This result is related to the point made by
Mullainathan et al. (2012) that underutilization of medical
care due to present bias can lower the moral-hazard cost
of providing insurance, potentially benefitting the insured.
16 See also Santos-Pinto (2008) for a closely related
analysis that focuses on identifying conditions under which
agent overconfidence benefits the principal in a singleprincipal setting. If effort is observable—so that only the
participation constraint is relevant—the principal benefits from agent overconfidence both because the agent is
overly willing to accept a given contract and because he can
exploit her as above. If effort is unobservable, the principal
benefits if the agent is control optimistic (and other, more
technical, conditions hold) as control optimism makes it
easier to satisfy the agent’s incentive-compatibility conditions as well.
1094
Journal of Economic Literature, Vol. LII (December 2014)
3.5 Potential Future Directions
In this section, I discuss some outstanding
issues in the literature in the same order in
which the related research appears above.
Further progress in the literature exploring
optimal moral-hazard contracts with social
preferences seems somewhat hindered by
the lack of a theory of reference-group formation—the determination of which other
individuals’ outcomes an agent cares about.
This not only makes it more difficult to test
or judge the economic importance of the
above results, but also prevents researchers
from asking natural questions, such as how
reorganizations affect the reference group
and hence behavior, in a theoretically disciplined manner.
While existing models with inequity-averse
decisionmakers can explain gift exchange in
the laboratory and in some organizations,
they do not predict the phenomenon when—
as in many real-life settings—the employer is
significantly wealthier than most employees
and cares mostly about his own material outcomes. Because in this case a generous wage
does not put the worker ahead, gift exchange
should not result. And because a selfish principal does not pay a bonus she does not have
to, the voluntary-bonus system should not
work.
Deepening the above puzzle, Netzer and
Schmutzler (2012) show that another major
model of social preferences, Rabin’s (1993)
model of intentions-based reciprocity, cannot convincingly account for gift-exchange
behavior, either, so that to date there is no
compelling explanation of this simple phenomenon. At first glance, Rabin’s model
appears to predict gift exchange: offering a
high wage seems to be a kind act, to which
the agent should naturally respond with her
own kind act of a high effort. In contrast,
Netzer and Schmutzler show that if the principal is selfish, this mechanism cannot work
to increase effort. By definition, a selfish
player never sacrifices his own payoff for that
of the other player, so the principal’s choice
of wage will never be considered kind by the
agent. As a result, the agent never exceeds
the materially cost-minimizing level of performance.17 This means that gift exchange
is either not about reciprocity, or a different
or more complex reciprocity mechanism is
involved. A model with agent heterogeneity
similar to that of von Siemens (2011b), discussed above, might provide a starting point,
but whether such a model will prove to be a
compelling general explanation requires further research.
The literature on loss aversion does not
seem to face similarly central puzzles, but
there are also plenty of unanswered questions. For instance, note that the contract
terms for loss-averse agents above emerge
because the firm offers partial insurance to
an agent who is first-order averse to risk.
From this perspective, it is interesting to note
that (as far as I am aware) no paper has carefully incorporated loss aversion into an insurance model with moral hazard and derived
implications for optimal contracts. The logic
of Herweg, Müller, and Weinschenk (2010)
seems to imply that in many settings, an
insurance contract with a simple deductible—whereby the agent pays a deductible if
losses exceed a threshold—is optimal.
While the literature on the interaction
between extrinsic and intrinsic motivation is
one of the most exciting and productive in
behavioral contract theory, it also calls for
substantial future research on when different sources of motivation are likely to dominate behavior. As explained above, all of the
17 Nevertheless, if the agent could choose lower levels of performance—that is, exert costly effort to punish
the principal—her wage can be much higher than that of
a selfish agent in the same situation, materially benefiting her and hurting the principal. In this case, the principal chooses a high wage not to induce discretionary high
effort, but merely to avoid the punishment that lower
wages would trigger.
Kőszegi: Behavioral Contract Theory
existing models explain only a part of the
evidence and make orthogonal or opposing
predictions on the other evidence—with no
sound guidelines as to where exactly each
model applies. One possible reason for the
confusing state of the literature may be that,
unlike extrinsic motivation, intrinsic motivation is a complex multifaceted phenomenon that is poorly understood. Individuals
may derive intrinsic motivation from many
sources, including those in the models
above, and perhaps additional motives such
as the sense of being in a good organization
or the sense of doing good work. Both more
evidence and more theory seems necessary
to sort out what situations evoke the different kinds of intrinsic motivation identified by
researchers.
The observation that a bias—here, slight
overconfidence—can benefit the agent raises
the natural question of when this is likely to
happen. Clearly, fixing the contract the agent
signs, a bias always (weakly) reduces her welfare. Hence, for a bias to benefit the agent, it
must change contract terms in her favor. This
can happen if the agent’s bias leads her to
either overvalue her outside option or undervalue interacting with the principal—both
forcing the principal to offer better terms.
The former is the case above, where competing principals offer terms the agent overvalues. The latter seems less likely to happen
in many settings, as the principal—in an
attempt to induce the agent to accept—will
try to write contracts the agent overvalues
rather than undervalues. But the question
of when biases can benefit an agent has not
received sufficient attention in the literature. Research on this question might draw
insights from the large classical literatures on
commitment and reputation, which also recognizes that appearing nonrational can benefit a player. For instance, the beneficial effect
of overconfidence above is similar to that of a
commitment to work hard, although the particular type of cost overconfidence implied
1095
has, to my knowledge, not been studied in
the classical literature.
4. Asymmetric Information and Screening
In classical screening models, a principal
(e.g., an airline) faces agents whose preferences (e.g., willingness to pay for a ticket) he
does not know. Ideally, the principal would
like to contract with multiple agent types,
differentiating the contract (e.g., the price
and conditions of the airline ticket) according to the agent’s preferences. The principal’s
main constraint—captured in the incentivecompatibility constraint that a more profitable type not take a less profitable option—is
that the agent may not reveal that she is a
profitable type (e.g., that she has a high value
for a ticket). The central issue in contract
design is how to trade off minimizing this
informational advantage with the objective
of achieving gains from trade. This section
summarizes the bulk of the literature incorporating ideas from psychology and economics into situations of hidden information and
screening.18
4.1 Commitment versus Flexibility
It has been understood that individuals with self-control problems benefit from
commitment, and hence might be willing to
sign contracts that restrict their choices in
some way. At the same time, uncertainty can
make it inefficient to remove all choice, thus
generating a possible tradeoff between commitment and flexibility. A number of papers
explore aspects of this contracting question.
In an early contribution, DellaVigna and
Malmendier (2004) study the optimal twopart tariff of a firm facing a consumer who
has known present-biased preferences and
18 Section 6 discusses a different set of screening issues
related to exploitative contracts, which arise because different agents might exhibit the mistake the principal is
looking to exploit to different extents.
1096
Journal of Economic Literature, Vol. LII (December 2014)
can make an investment such as saving or
exercising. Investment carries an immediate
cost that is unknown to the firm, and generates a fixed future benefit that is known.
If the consumer is sophisticated regarding
her present bias, the optimal two-part tariff
implements the first-best outcome by imposing a subsidy that equals the amount by
which the consumer undervalues the future
benefit. Hence, the contract induces more
future-oriented behavior without compromising the consumer’s flexibility to respond
to cost shocks. While this contract achieves
first-best despite asymmetric information
about the cost, it is clear that the optimality
is specific to the problem and form of uncertainty: for instance, if the consumer’s benefit
was also uncertain, the first-best could not be
achieved with a two-part tariff.
Amador, Werning, and Angeletos (2006)
study optimal contracting in a setting where
the first-best is typically not achievable.
They consider a consumption–savings problem in which a present-biased agent faces a
privately observed taste or needs shock that
affects her marginal instantaneous utility of
consumption and that she does not initially
know.19 Thinking of the consumer’s late self
as the agent and her early self (or a social
planner maximizing the expected utility of
her early self) as the principal, they characterize the optimal commitment, defined
as the optimal subset of the agent’s budget
set from which she will be allowed to choose
her savings level. Amador et al. show that the
optimal commitment always features a minimum savings rule akin to those observed in
many retirement systems. Intuitively, it is not
optimal to allow the highest types—those
with the greatest taste for immediate consumption—to choose higher consumption
19 Technically, this kind of situation is called a “hidden
knowledge” problem, because—as distinct from non-linear
pricing or screening problems—exogenous asymmetric
information emerges after the contract is signed.
levels than slightly lower types, as this would
mean that the highest types are overconsuming from an ex ante point of view. In addition,
Amador et al. show that it is often optimal to
allow for complete flexibility above the minimum savings requirement. Roughly, whatever flexibility in savings the principal allows
above the minimum, the agent tends to overconsume in that range from an ex ante point
of view. Since the principal cannot do much
about the overconsumption, he might as well
allow the agent to adjust consumption to the
taste shock.20
Galperti (2012) studies a setting where,
much like in DellaVigna and Malmendier
(2004), the principal can implement firstbest if she knows the agent’s degree of time
inconsistency, but a screening issue arises
because there are both time-inconsistent and
time-consistent agents in the population. A
flexible contract that solves the time-inconsistent agent’s commitment problem must
offer rewards for saving. But because a timeconsistent agent can expect to receive the
rewards more often, she derives rent from
such a flexible contract. To lower the timeconsistent agent’s information rent, then, the
principal curtails the flexibility of the timeinconsistent agent’s contract, restricting
both high and low levels of savings. Galperti
argues that this feature is consistent with
restrictions on retirement savings devices in
the US.
Esteban, Miyagawa, and Shum (2007)
study a monopolist’s nonlinear pricing
problem in which the heterogeneity in
20 This insight does not always hold due to a subtle
consideration in the principal’s problem. By eliminating a
range of possible consumption levels, the principal forces
those types who would otherwise have preferred to consume in that range to consume either less or more. The
former response is welfare-increasing, while the latter
response is welfare-decreasing. The former force dominates the combined harmful effects of the latter force
and lowers flexibility if sufficiently more types respond by
consuming less, which is the case if the density of types
decreases sufficiently quickly.
Kőszegi: Behavioral Contract Theory
agent p
references is due to heterogeneity in temptation disutility in the sense of
Gul and Pesendorfer (2001).21 In Gul and
Pesendorfer’s model, the agent has preferences consisting of commitment utility—the
utility of an option if the agent can perfectly
commit to it in advance—and temptation
disutility, where she suffers the latter if she
does not choose the most tempting option
from her choice set. Whereas with hyperbolic discounting the agent would like to
commit her future behavior because she
will have different preferences in the future
than she does now, with temptation disutility the same demand for commitment arises
because she wishes to lower temptations. The
impact of temptation disutility on the optimal menu turns out to depend on the kind
of temptation agents experience. If all agents
experience upward temptation—whereby
high consumption is more tempting than
low consumption (e.g., cigarettes)—then the
optimal menu is a singleton, so that the seller
does not separate consumers with different
preferences at all. Intuitively, because trying
to take advantage of an agent’s temptation
to sell her more would make her less willing to participate, the seller just chooses to
maximize agents’ commitment utility, which
is possible with a singleton since (by assumption) agents have homogenous commitment
utility. If all agents experience downward
temptation—whereby lower consumption
is more tempting than higher consumption
(e.g., exercise)—the optimal menu is identical to that with standard preferences equal
to ex post preferences. Intuitively, since not
buying is the most tempting alternative and
consumers can always choose this alternative, increasing the menu does not increase
temptation disutility, so that it does not make
consumers less willing to participate. Hence,
the seller screens consumers
according
21 Esteban and Miyagawa (2006) analyze the competitive case.
1097
to ex post preferences. The logic of these
results (especially those for upward temptation) relies crucially on the assumption that
individuals differ only in temptation disutility, and as the authors point out, it would
be interesting to study the same problem
if commitment utility also differs between
types.
4.2 Overconfidence and Screening Issues
Related to False Beliefs
The observation that some individuals
have overly positive views about themselves
or their prospects raises the question of how
to optimally screen agents with different
levels of overconfidence. Before discussing
research on this question, it is worth noting
a general limit to screening beliefs according
to their accuracy: because two individuals
with the same beliefs about future outcomes
and same ex ante preferences choose from a
menu in the same way, separating them at
the contracting stage by means of self-selection is impossible. In particular, this is the
case even if one has correct and the other
incorrect beliefs, and hence the two agents
will behave differently given the contract.
As a result, a contract signed by agents with
given beliefs are often priced according to
some average of the actual outcomes of these
agents. In these situations, a biased agent
exerts an externality on rational agents. The
externality is negative if the biased agent is
less profitable than the rational agent (e.g.,
in insurance, where an overconfident agent
underestimates her risk), and positive if the
biased agent is more profitable (e.g., in many
exploitative contracts discussed below).
Overconfidence in the insurance market.
Sandroni and Squintani (2010) study insurance contracts when some high-risk agents
believe that they are low-risk, and show that
the presence of overconfident agents has
qualitative observable implications in a competitive (but not in a monopolistic) insurance
1098
Journal of Economic Literature, Vol. LII (December 2014)
market. To understand their results, recall
that in the seminal competitive insurance
model of Rothschild and Stiglitz (1976) with
rational agents, high-risk and low-risk agents
separate in equilibrium and the price of insurance is actuarially fair for both types, so that
the price of insurance in each class does not
depend on the proportion of types in the population. In addition, because high-risk agents
able to fully insure at actuarially fair prices
do not want small amounts of insurance even
at low prices, low-risk agents receive at least
partial insurance. Overconfidence changes
each of these predictions. If some agents
who believe themselves to be low-risk are in
fact high-risk, the group of low-risk and overconfident—who by the above logic cannot be
distinguished because they choose contracts
in the same way—receive more expensive
insurance that depends on the proportion of
overconfident in the population, generating
heterogeneity in insurance prices within a
risk class. In addition, since the low-risk and
overconfident believe that their insurance is
overpriced, they may not buy insurance.
Sandroni and Squintani (2007) reconsider
the scope for Pareto-improving interventions
in the insurance market when overconfident
agents are present. Although in a rational
competitive insurance market low-risk agents
receive cheap insurance, such insurance is
partial to discourage high-risk agents from
taking it. In this equilibrium, low-risk agents
would prefer to buy more insurance at the
same price, so that a government policy of
mandatory insurance—which allows low-risk
agents to get more insurance at reasonable
prices—can be Pareto-improving. But with
overconfident agents, the group of low-risk
and overconfident are offered more expensive insurance, and since this group believes
that they are receiving insurance above the
actuarially fair price, they may not want
much insurance. If this is the case, the group
of low-risk and overconfident can choose a
perceived-optimal insurance contract that is
rejected by high-risk consumers. As a result,
low-risk and overconfident consumers prefer not to buy more insurance at prevailing
prices. Hence, contrary to common intuition
that biases increase the scope for government intervention, in this case mandatory
insurance is not Pareto-improving.22
Contrary to the prediction of classical
insurance models that higher-risk consumers buy more insurance, recent empirical
research finds a zero or negative correlation between risk and insurance coverage in
many circumstances (for instance Chiappori
and Salanie 2000; Finkelstein and McGarry
2006). To explain this puzzle, Spinnewijn
(2013) studies insurance markets with optimistic agents and both moral hazard and
asymmetric information. If a single-crossing
property holds, the only way for a principal
to separate agents is to offer them different amounts of insurance, and this separates
agents (roughly) only according to baseline
optimism. But because control optimism
affects how much precaution an agent takes
when insured, the correlation between risk
and insurance coverage depends on the correlation between baseline optimism and control optimism.
Overconfidence and debt financing.
Manove and Padilla (1999) consider a model
in which overoptimistic entrepreneurs
ask for loans from banks to finance overly
large projects relative to the most productive use of money, and banks cannot distinguish overoptimistic entrepreneurs from
realists. Manove and Padilla’s main result
22 Schumacher (2012) studies a model with a related
effect, where—similarly to the overconfident in Sandroni
and Squintani (2007)—naïve present-biased agents exert a
negative pricing externality on sophisticated agents by not
taking enough precautions, making government intervention less desirable. Nevertheless, in Schumacher’s (2012)
model, government intervention to induce taking more
precaution is Pareto-improving if the share of naïve agents
is substantial.
Kőszegi: Behavioral Contract Theory
is that banks are often willing to fund too
many projects, from a social point of view.
Intuitively, banks care only about recouping
their investment in expectation, and do not
internalize the fact that an optimistic entrepreneur could have used the money better
elsewhere. Furthermore, because collateral
requirements help ensure that banks recoup
their loans, they exacerbate the excessive
lending problem.
4.3 Other Topics in Screening
This section discusses various other
research on screening.
Using framing for screening. Salant and
Siegel (2013) consider the screening problem of a seller who can temporarily manipulate buyers’ preferences. The seller chooses
quality and price for two types of buyers,
and in addition may employ a “frame”—such
as a salient comparison price in a shop or a
glitzy environment in a casino—to change
the types’ utility functions. Crucially, however, the effect of the frame is fleeting, and
either (i) the buyer returns the item if she
later realizes that she does not value it above
her outside option, or (ii) the buyer anticipates the manipulation and stays away from
the store if she does not like what she will
buy. Salant and Siegel show, roughly, that
such framing can help the seller if and only if
it relaxes the high type’s incentive constraint
without increasing the low type’s incentive to
mimic the high type. In this case, the principal can offer the low type a cheap product
while manipulating the high type into buying
the expensive product, eliminating the high
type’s information rent and increasing efficiency of provision to the low type.
Sorting in the labor market. Kosfeld and
von Siemens (2011), von Siemens (2011a),
and von Siemens (2012) study labor-market
outcomes for employees with different types
of social preferences. In the monopsony
1099
model of von Siemens (2011a), employees might be selfish or inequity-averse with
respect to coworkers. Since a worker can
affect only her own outcome and both selfish and inequity-averse workers prefer more
money to less (holding others’ outcomes
constant), the firm cannot differentiate selfish and inequity-averse workers who have
been hired. In order to employ low-ability
inequity-averse individuals, therefore, a
firm needs to pay a premium to compensate
these workers for falling behind coworkers.
To economize on this cost, the firm either
reduces the gap between high- and lowability workers by distorting production, or
excludes low-ability inequity-averse workers by paying no premium. In contrast, von
Siemens (2012) shows that in a competitive market, the existence of multiple firms
allows high- and low-ability workers to sort
into different firms, reducing the impact of
social comparisons, and hence often allowing low-ability inequity-averse workers to
be employed as well. In a related setting,
Kosfeld and von Siemens (2011) assume
that some workers are selfish—i.e., exert
individual effort if compensated for it, but
never exert team effort—while other workers are conditional cooperators—i.e., exert
team effort if coworkers do too, and prefer to
work in a team environment. In equilibrium,
firms offer two types of contracts: a highpowered incentive contract designed for the
selfish workers, and a low-powered incentive contract designed for the conditional
cooperators. To prevent selfish workers from
accepting the latter contract, its wage must
be relatively low, so that it may be profitable
for the firm despite competition. This means
that if team effort is sufficiently important,
firms with lower pay are more productive.
Loss aversion and screening. Hahn et al.
(2010) analyze a monopolist’s optimal menu
when consumers are loss averse and do not
know their willingness to pay in advance. Just
1100
Journal of Economic Literature, Vol. LII (December 2014)
as a loss-averse employee dislikes the risk
inherent in a wage schedule that discriminates finely between performance levels (see
section 3.2), a consumer who does not initially know her willingness to pay dislikes the
risk inherent in a menu that discriminates
finely between different willingness-to-pay
realizations. Hence, the seller often offers a
small number of products relative to the heterogeneity in the population.
Carbajal and Ely (2012) assume instead
that consumers know their types in advance
and have a type-dependent reference point
relative to which they evaluate outcomes.
Carbajal and Ely study how the optimal
menu depends on the reference-point
function, and also derive properties of selfconfirming reference consumption plans—
where a type’s consumption in equilibrium
coincides with her reference point. In contrast to an individual-decision-making setting—where an increase in the reference
point always hurts the agent—a higher selfconfirming reference consumption plan can
benefit both the seller and some agents.
Intuitively, a higher reference point leads the
seller to exclude fewer low types from the
market (who, due to their higher reference
point, value the product more highly), and
as a result of this market expansion, higher
types receive higher information rents.
4.4 Potential Future Directions
Screening seems to be an understudied
topic in behavioral contract theory. Notably,
behavioral research has not been incorporated into optimal income taxation, although
some psychological phenomena (such as
hyperbolic discounting, anticipatory utility, and overconfidence) seem important
in labor, leisure, and consumption choices.
More generally, there seems to be little
research on how insights from psychology
and economics affect classical screening
problems in which the private information
concerns a standard preference parameter;
in much of the research above, the contract
does not have a classical purpose or the private information concerns a parameter in
psychology and economics.
Another important question is the extent
to which a social planner can screen agents—
some of whom might be behaving optimally,
and some suboptimally—in the context of
noncoercive interventions aimed at improving individual consumer decisions.23 For
instance, O’Donoghue and Rabin (2003) illustrate that a “sin license,” whereby consumers
would be required to buy a moderately priced
license to smoke, often dominates “sin taxes”
because it separates time-consistent consumers from both naïve and sophisticated present-biased consumers. Intuitively, neither
type of p
resent-biased consumer obtains the
license—naïve consumers because they do
not believe they will smoke in the future, and
sophisticated consumers because they want
to prevent their future selves from smoking—
but time-consistent consumers who are making an optimal decision to smoke do.
5. Auction Theory and Mechanism Design
Mechanism design is the study of what
outcomes can be achieved, and how, when
a principal is interacting with multiple
agents with private information. The central
consideration in the principal’s problem is
how to set up the game form so that agents
reveal their private information. This section
reviews applications of behavioral-economics ideas to mechanism design. The bulk of
existing research concerns the theory of auctions, a prominent class of mechanisms for
selling to buyers with unknown valuations
for the product. The section also describes
the limited amount of work on other issues
in mechanism design.
23 See Thaler and Sunstein (2008) and Camerer et al.
(2003) for arguments in favor of such interventions, as well
as some examples.
Kőszegi: Behavioral Contract Theory
5.1 Loss Aversion in Auctions
When applying loss aversion to auction
theory, predictions depend on whether a bidder experiences gain–loss utility separately
from money and from the product being sold
(separate evaluation), or jointly from the net
utility of the transaction (net evaluation). In
commodity auctions—where the payment
is monetary and the product is nonmonetary—the former assumption seems more
appropriate, while in induced-value laboratory experiments—where both the payment
and the “product” are monetary—the latter
assumption seems to apply.24 As Lange and
Ratan (2010) argue, this implies both that
one must use different theories to interpret
induced-value laboratory auctions and commodity auctions, and that many experimental
findings may not apply to real-world settings.25
Lange and Ratan (2010), Eisenhuth
(2010), and Eisenhuth and Ewers (2012)
analyze the effect of expectations-based loss
aversion as modeled in Kőszegi and Rabin
(2007) on the revenue ranking of standard
sealed-bid private-values auctions. Kőszegi
and Rabin’s model implies that individuals
are first-order averse to local risk, and hence
bid more aggressively in auction formats
that are better at insuring them against the
uncertainty from participation. This behavior
24 The perspective that individuals use separate evaluation when a real commodity is traded off with money
is consistent with a large body of evidence, such as the
endowment effect, that is commonly interpreted in terms
of loss aversion. Nevertheless, Eisenhuth and Ewers
(2012) present some suggestive experimental evidence
that a model with net evaluation better describes a commodity auction, as well.
25 Lange and Ratan (2010) find that with net evaluation,
bidders “overbid” in a first-price private-values action,
while with separate evaluation they underbid. Intuitively,
with net evaluation, loss-averse agents experience a sensation of loss compared to the expected payoff when
losing the auction, motivating them to bid higher. In contrast, with separate evaluation, loss aversion in the money
dimension kicks in when the agent wins the auction, and
thus results in agents underbidding in order to avoid having to pay more than expected.
1101
generates a revenue ranking between the
three most commonly analyzed auction formats—first-price, second-price and all-pay
auctions—that depends on whether bidders
use separate or net evaluation. Under separate evaluation, an all-pay auction is less risky
than a first-price auction because payment is
deterministic, and a first-price auction is less
risky than a second-price auction because
payment conditional on winning is deterministic. In fact, Eisenhuth (2013) establishes that under loss aversion with separate
evaluation, any optimal mechanism features
a riskless payment, and the optimal auction
is an all-pay auction with a minimum bid.
In contrast, with net evaluation the ranking
between the all-pay and first-price auctions
is reversed. Intuitively, the gap between the
net utility from winning and not winning
equals the product’s value with an all-pay
auction but only the surplus with a first-price
auction, so the latter appears less risky for
a net evaluator. In fact, Eisenhuth (2010)
proves that in this case, the optimal auction is
the first-price auction with a minimum bid.26
Another possibility explored by researchers is that the reference point of auction
participants is partially determined by the
reserve price. In particular, Rosenkranz and
Schmitz (2007) specify the reference point in
money as a weighted average of the reserve
price and an exogenous parameter, so that an
increase in the reserve price directly raises
bidders’ willingness to pay, and hence their
bids. When either the number of bidders
or the exogenous component of the reference point is high, the bid-raising benefit of
increasing the reserve price dwarfs the cost
from the increased risk of not being able to
sell, and therefore even a small degree of loss
26 From a theoretical perspective, a weakness of the
above literature is that (as the authors note) it makes predictions qualitatively similar to those of appropriately specified models with classical risk-averse bidders. Of course,
the implications of loss aversion could be quantitatively
more important.
1102
Journal of Economic Literature, Vol. LII (December 2014)
aversion can lead to a significant increase in
the optimal reserve price. Shunda (2009)
assumes that bidders’ reference points
depend on the seller’s reserve price as well
as his “buy price”—a price at which a bidder may purchase the good from the seller
before the auction starts. Increasing the buy
price raises the bidders’ willingness to pay,
and hence leads to a larger pool of participants and higher bids in the auction stage,
but lowers the probability that a bidder buys
the item in the first stage. The optimal buy
price trades off these two effects.
5.2 Other Topics in Auctions
Nonequilibrium thinking. Jehiel and Lamy
(2012) ask why sellers often employ absolute auctions (auctions with a reserve price
of zero) despite positive value for the object,
and why sellers often set secret reserve prices.
In the spirit of Jehiel (2005) and Eyster and
Rabin (2005), Jehiel and Lamy propose that
some bidders do not understand how the
potential for getting a good deal varies with
the auction format. Absolute auctions can be
used to attract bidders who underappreciate
how product quality and the participation
rate depend on the auction format, as these
bidders overestimate the quality of goods in
such auctions and may also underestimate
the participation rate. And auctions with a
secret reserve price can be used to attract
bidders who do not understand how secret
reserve prices differ in distribution from public reserve prices, failing to anticipate that a
secret reserve price is likely to be higher than
public reserve prices.
Augenblick (2011) studies the penny auction, an auction format in which agents bid for
items in predefined increments (often 1 cent)
and have to pay a nonrefundable fixed price
for each bid. Augenblick
documents that
bidders severely overbid in these auctions,
generating for instance an average profit margin of 104 percent in auctions for direct cash
payments. He argues that the penny auction
engages a naïve sunk-cost fallacy, whereby
bidders are reluctant to stop bidding in an
auction when they have spent money on it,
and do not predict the full extent of this effect.
He shows that the profit-maximizing supply of
auctions is constrained by the consideration
that an auction attracting few bidders tends
to end early and generate large losses for the
auctioneer. The same consideration also creates an effective barrier to entry, as competitors who cannot initially attract many bidders
must absorb large losses.
Crawford et al. (2009) begin studying
optimal auction design with level-k bidders.
Following Camerer, Ho, and Chong (2004)
and Crawford and Iriberri (2007), their
model distinguishes between bidders of different levels of strategic reasoning. L0 types
follow some simple bidding strategy, such as
bidding a random number or always bidding
their true valuation. L1 types formulate their
strategy as the best response to the L0 strategy. For example, if L0 types bid randomly,
then L1 types believe that winning does not
reveal any information about the value of the
object, so that they—similarly to fully cursed
players in Eyster and Rabin (2005)—overbid
compared to the predictions of the classical
model. L2 types best respond to L1 types.
Crawford et al. (2009) find that the optimal
reserve price may either be higher or lower
than in an equilibrium model, while seller revenue—due to overbidding by many players—
is usually higher. They also give an example
of an exotic auction specifically designed to
exploit bidders’ nonequilibrium thinking that
can yield arbitrarily large revenues.
Anticipated regret. Engelbrecht-Wiggans
(1989) and Filiz-Ozbay and Ozbay (2007)
study a model in which bidders anticipate
feelings of regret in case their bid turns out to
be suboptimal ex post. Losers of a first-price
sealed-bid auction or a Dutch auction may
feel regret when confronted with the information that the winning bid was lower than
Kőszegi: Behavioral Contract Theory
their valuation, which means that they could
have walked away with a positive surplus had
they followed a more aggressive strategy. On
the other hand, winners of a first-price auction
may also feel regret when discovering that
the second-highest bid was lower than their
own, so that they could have won at a lower
price. Anticipating “loser regret” induces participants to make higher bids than otherwise,
while anticipating “winner regret” induces
them to make lower bids. These emotions
are only triggered when the regret-inducing
information is revealed to participants, so a
seller can influence bidder behavior through
information provision. In a setting that triggers only loser regret, participants overbid
relative to their consumption utility.27
5.3 Mechanism Design
Beyond auction theory, there is little work
on the psychology and economics of mechanism design. One exception concerns the
implications of social preferences in publicinterest situations, those in which everyone
benefits if everyone gives up personal material
gain for the sake of others. In a number of different models, authors find that social preferences tend to increase efficiency. Kucuksenel
(2012) considers mechanism design with
altruistic agents and shows that more altruistic agents—those who attach a larger weight
to others’ material payoffs—are more likely
to produce a public good that is efficient to
produce, and are more likely to trade when
it is efficient to do so. The intuition is simple:
more altruistic agents care more about social
efficiency, so it is easier to implement socially
efficient outcomes with them. Bierbrauer and
Netzer (2012) establish a similar result for
27 See also Cramton et al. (2012) for an application of
this concept to clock auctions such as those commonly
used for selling radio spectrums, electricity, or gas. And
Roider and Schmitz (2012) study optimal reserve prices
in a related model that assumes bidders get utility directly
from winning or losing, independently of whether they
could have done better with another strategy.
1103
agents with intentions-based social preferences in the sense of Rabin (1993), where a
player is willing to sacrifice her own material
payoff for the sake of the other player if and
only if she believes the other player is similarly kind. The codependence of intentions
makes kindness an equilibrium phenomenon. If in equilibrium agents have positive
intentions toward each other, then the observation that they can achieve better outcomes
is similar to that in Kucuksenel (2012). But
since intentions are endogenous, a key part
of the designer’s problem is to set up a mechanism in which agents can develop positive
intentions toward each other. Bierbrauer and
Netzer show that the designer can achieve
this by adding unused options to enrich oneself to the mechanism, so that truth-telling
will be considered a kind action. Even so,
there is always an equilibrium in which players are unkind to each other, so the designer
must make sure that the kind equilibrium is
played—for which the theory provides no
guidance. Finally, Carmichael and MacLeod
(2003) make a related point in the context
of holdup: they show that caring about the
other party’s investment—a notion akin to
a preference for equity—can soften the
holdup problem in contracting.28, 29
28 Desiraju and Sappington (2007) consider a nonlinear
pricing problem with two agents who are averse to inequity in the total ex post payoff they receive, a setting that
does not neatly fit the public-interest environment above.
While with standard preferences a two-agent screening
problem reduces to a single-agent screening problem, with
inequity aversion it is fundamentally multi-agent mechanism design, as each agent cares about the other’s surplus.
Desiraju and Sappington show that if the agents are ex ante
identical, the principal can implement the standard outcomes while eliminating all ex post inequality. In one such
mechanism, a low-type agent gets paid more if the other
agent is a high type than if the other agent is also a low
type. If the agents are ex ante different, however, there is
no way to implement the standard outcomes while eliminating inequality for all type realizations.
29 There is a also a mini-literature on implementation when agents have a preference for reporting their
types honestly, showing that even a weak such preference
can greatly help the principal (Alger and Renault 2006;
1104
Journal of Economic Literature, Vol. LII (December 2014)
5.4 Potential Future Directions
Several topics in mechanism design are
natural candidates to link to behavioral
ideas. A notable aspect of auctions is that
they seem to trigger some unique emotions
related to competition that are not triggered in many other settings. In particular,
as noted by Malmendier and Lee (2011),
it has been understood since ancient times
that auctions invoke “bidding fever,” while
Morgan, Steiglitz, and Reis (2003) and Cox,
Smith, and Walker (1988) argue that auctions engage a spite motive and the joy of
winning, respectively. An interesting issue
is how these emotions work and how they
affect optimal auction design. For instance,
it is not immediately clear why these emotions are triggered by auctions, while other
situations (including many in this review)
are more likely to generate positive social
preferences.
While classical mechanism design recognizes that public revelation of information
can help the principal—e.g., in an auction if
bidders’ valuations are positively correlated
(Milgrom and Weber 1982)—behavioral
economics introduces at least two novel reasons for why information could matter. First,
as in the case of regret above and in the case
of anticipatory utility discussed in other sections of the review, information can directly
affect agents’ preferences. Second, information can lead strategically naïve agents
to change their views about the strategic
situation, either directly through providing information on how others play or indirectly through focusing their attention on
specific aspects of the game. For instance,
in a second-price auction a bidder must rely
exclusively on her predictions regarding the
Matsushima 2008). These papers, however, make specific
assumptions about the structure of honesty preferences
that do not seem to be based on psychology evidence, so I
do not review them here.
distribution of bids, but in an English auction she receives some information on this
distribution. Such informational issues seem
understudied in auction theory and mechanism design.
6. Exploitative Contracting
This section considers contracts designed
exclusively or primarily to exploit false beliefs
by the agent. While there does not seem to
be a precise formal way to distinguish such
contracts from nonexploitative contracts that
are affected by agent mistakes, an informal
and subjective distinction seems useful to
make. Namely, a contract is exploitative if
the economically central considerations driving it derive from trying to profit from the
agent’s mistake, and other considerations or
constraints are nonexistent, not binding, or
not central.
The literature has explored two broad
forms of false beliefs. First, an agent’s false
beliefs may be about the contract itself; in
Gabaix and Laibson (2006), for instance,
naïve consumers underappreciate that the
product they are purchasing will require
some add-on purchases. Alternatively, an
agent might believe that regulation prevents
certain charges, when in fact it does not
(Armstrong and Vickers, 2012). Second, an
agent’s false beliefs may be about her own
behavior given the contract; in DellaVigna
and Malmendier (2004), for instance, naïve
hyperbolic discounters are aware of all contract features, but they mispredict how they
will behave given a contract.
Methodological issues. The models discussed in this section assume that the principal knows the agent’s tendency to commit
mistakes, so that in this domain, the principal is more informed about the agent than
she is herself. While unusual from a classical point of view—where it is commonly
assumed that the agent’s tendencies are
Kőszegi: Behavioral Contract Theory
her private information—this assumption
often makes sense when the principal is a
firm with the capacity and willingness to
collect and analyze tremendous amounts
of data about consumers, and the agent
is an individual consumer. The theories
are also unlike informed-principal models
with novel types of principal information
because the agent does not make inferences about her tendencies from the contract she is offered. Hence, as discussed
by Eliaz and Spiegler (2006), these models
can be treated formally as contracting with
noncommon priors. While the classical literature also recognizes the potential relevance of noncommon priors for contracting
and other settings, the recent literature has
both reinvigorated research on such models
and changed its style and emphasis: the theories posit a particular form of agent beliefs
based on psychology evidence, take a stance
on which beliefs are correct, and tend to
consider welfare issues.
6.1 Nature and Implications of Exploitative
Contracts
Seemingly cheap products. Exploitative
contracts often make products appear
cheaper than they really are. The reason is
simple: firms want to encourage consumers
to buy, so if they choose multiple prices (e.g.,
a basic price and an add-on price) that consumers will pay, they aim to obtain revenues
more from the prices consumers underappreciate. In a competitive market, this does
not necessarily affect firm profits or consumer or social welfare: similar to the logic
of loss-leader pricing (Lal and Matutes 1994)
as well as that of many switching-cost models (e.g., Farrell and Klemperer 2007), firms
compete aggressively for valuable consumers
ex ante and bid down the more noticeable
component of the price until they eliminate
net profits. In this sense, naïve consumers
can be protected from their mistakes by market forces.
1105
Nevertheless, the fact that firms distort
prices to take advantage of consumer mistakes can have a number of efficiency implications, so that the above “safety in markets”
is very partial. Consumers might buy products whose value is below the true price but
above the misperceived low price, generating inefficiency in participation decisions
(Heidhues and Kőszegi 2013). Other inefficiencies obtain due to the high prices naïve
consumers do not appreciate. In Gabaix and
Laibson (2006) and Armstrong and Vickers
(2012), naïve consumers ignore add-on
prices, firms respond by choosing high addon prices, and this leads sophisticated consumers to take socially inefficient steps (such
as arranging for a substitute with higher production costs) to avoid the add-on. In Grubb
(2009), consumers believe they can predict
their consumption more precisely than they
actually can, firms respond by setting high
marginal prices for high usage, and this can
lead consumers to underutilize the service
ex post. In Gottlieb and Smetters (2012), lifeinsurance buyers do not properly account for
the probability that they will lapse their policies, firms respond by front-loading premiums to make the policy look better, and this
distorts consumption smoothing. At the same
time, distorted prices could also have beneficial effects if they correct another problem. For instance, firms take advantage of
naïve present-biased agents by offering high
per-usage fees for pleasurable goods whose
consumption the agent underestimates, and
these high prices have the beneficial effect
of lowering the present-biased agent’s consumption (DellaVigna and Malmendier,
2004). Further, as argued by Bar-Gill (2006),
if a consumer’s mistake leads her to consume
too little, the higher consumption induced
by deceptively low prices can benefit her.
Cross-subsidies. Another recurring feature of exploitative contracts, first discussed
in detail by Gabaix and Laibson (2006), is
1106
Journal of Economic Literature, Vol. LII (December 2014)
that sophisticated consumers benefit from
the presence of naïve consumers. Because
naïve consumers do not anticipate some
fees that they will pay, they tend to generate
higher profits than sophisticated consumers
buying the same contract. In a competitive
market—where firms make zero profits on
average—it must therefore be that firms
make money on naïve consumers and lose
money on sophisticated consumers, in effect
acting as a tool for cross-subsidizing the latter type.
Exacerbation of mistakes using high fees.
DellaVigna and Malmendier (2004) were the
first to point out that firms might fine-tune
contracts to exacerbate consumers’ mistakes.
A number of papers explore the specific features of such exploitative contracts.
Eliaz and Spiegler (2008) study optimal
dynamic contracting when an agent’s preferences are uncertain at the time of signing
the contract, and the agent may be more
optimistic than the principal about the better state occurring. As an example, suppose
that all law-school students will end up
in the nonprofit sector or in the corporate
sector with the same fixed probabilities,
but students have different, to the school
unobservable, overoptimistic beliefs that
they will end up in the nonprofit sector. The
school’s optimal screening menu includes a
standard nonexploitative loan contract that
is not contingent on subsequent outcomes,
as well as exploitative loan contracts that
amount to speculation on where a student
will end up. With the latter type of loan,
the student pays somewhat less if she ends
up in the nonprofit sector and much more
if she ends up in the corporate sector. An
interesting feature of the optimal menu is
that students with beliefs close to those of
the school receive a nonexploitative contract. Intuitively, although the school could
make a little money on slightly overoptimistic students by taking a small bet on what the
student will do, much more overoptimistic
students would value such a contract much
more highly, forcing the school to pay a kind
of information rent. This implies that contracts should either involve no speculation or
substantial levels of speculation.
Eliaz and Spiegler (2006) analyze a
closely related problem in which the principal knows that the agent’s preferences
will change between the time of signing the
contract and the time of taking an action,
whereas the agent may assign positive probability to her preferences remaining the same.
The time inconsistency in the agent’s preferences introduces an additional value from
contracting similar to that in section 4.1: the
agent prefers the contract to restrain her
future behavior. Again, if the agent’s beliefs
are sufficiently close to that of the principal,
she receives the ex ante optimal contract that
fully commits her behavior, but if she is sufficiently naïve regarding her time inconsistency, she receives an exploitative contract
in which she is effectively rewarded if she
does not change her mind and punished if
she does change her mind.
Heidhues and Kőszegi (2010) study exploitative credit contracts, and analyze the kinds
of mistakes that are exacerbated by firms and
lead to large welfare losses. In their model,
present-biased consumers who might be
partially naïve about their time inconsistency
can sign exclusive contracts with competitive
suppliers of credit, agreeing to a menu of
installment plans according to which credit
can be repaid in the future. In the competitive equilibrium firms offer seemingly cheap
credit to be repaid quickly, but introduce
large penalties for falling behind this frontloaded repayment schedule. The contracts
are designed so that borrowers with even an
arbitrarily small degree of naïveté both pay
the penalties and repay in an ex ante suboptimal back-loaded manner more often than
they predict. Intuitively, a lender chooses the
repayment options so that, when deciding
how to repay in the future, the consumer will
Kőszegi: Behavioral Contract Theory
be indifferent between the front-loaded and
back-loaded schedules. As a result, if she is
naïve about her time inconsistency—no matter by how little—she falsely believes that she
will repay quickly. To make matters worse,
the same misprediction leads the consumer
to underestimate the cost of credit and borrow too much—despite borrowing being for
future consumption.
6.2 Helping Consumers
Many researchers have asked whether
there are ways of improving consumer welfare in the types of situations described in
the previous subsection. The theme that
emerges from this literature so far is that both
market-based and regulatory solutions face
severe problems, but some forms of intervention may nevertheless increase welfare.
As has been pointed out by Armstrong and
Vickers (2012), for instance, one common
feature of welfare-increasing interventions
is redistribution of wealth from sophisticated
to naïve consumers through the reduction of
cross-subsidies. While some researchers and
observers use this as an argument against
intervention on the basis that we should not
hurt rational agents who are paying attention
and making individually optimal decisions,
from an economic perspective this seems no
better a reason to refrain from intervention
than the fact that pollution taxes redistribute
wealth from potential polluters to the rest
of society. In fact, to the extent that sophisticated consumers are wealthier than naïve
consumers, redistribution might be an additional argument for intervention.
Lack of market incentives to educate consumers. The fact that consumers underestimate some costs associated with product use
raises the question of whether market participants have incentives to educate consumers
or offer more transparent products or contracts. First-pass logic suggests that in a sufficiently competitive industry, competitors
1107
should unshroud price components underappreciated by consumers, and then compete
on them. While it is unclear whether and
how one can educate consumers, and this
question requires further research, a number of papers have investigated the incentive to unshroud assuming that firms can
costlessly do so. Gabaix and Laibson (2006),
Spiegler (2006), and Heidhues, Kőszegi, and
Murooka (2012) show that unshrouding is
often unprofitable because it turns profitable
naïve consumers into unprofitable sophisticated consumers. Hence, deceptive products
or contracts can often survive in markets.
Furthermore, Heidhues, Kőszegi, and
Murooka (2012) identify a perverse aspect of
when this can happen: products that generate
lower social surplus than the best alternative
facilitate deception precisely because they
would not survive in the market if consumers
understood hidden fees, and therefore firms
often make profits on exactly such products.
Problems with financial advice. Armstrong
and Zhou (2011), Stoughton, Wu, and Zechner
(2011) and Inderst and Ottaviani (2012) show
that if consumers are naïve, firms pay commissions to financial advisors to steer consumers
to buy their high-priced products, and financial advisors do so even if this is suboptimal
for a consumer. Murooka (2013) establishes
that even perfect competition among financial
advisors does not push commissions down to
a competitive level, as deceptive firms must
pay significantly higher commissions to stop
advisors from explaining to consumers that
other products are superior. Since the high
commissions are ultimately paid by consumers, in this situation the presence of financial
advisors who can explain deceptive practices
lowers consumer welfare.
Interventions to improve consumer decision making. A number of researchers recognize that when consumers might mispredict
their own behavior, classical disclosure, which
1108
Journal of Economic Literature, Vol. LII (December 2014)
typically limits information to those related to
the contract itself, is insufficient, and informing consumers about themselves is necessary
(for example Bar-Gill and Ferrari 2010). An
approach by Thaler and Sunstein (2008) and
Kamenica, Mullainathan, and Thaler (2011)
would require firms to provide contract details
and individualized usage information to all
consumers in a standard form, so that marketbased “choice engines” can emerge that help
consumers choose, making consumer choices
but not necessarily making consumers more
sophisticated.
Welfare-improving interventions that
limit contract forms. A number of authors
have shown that limiting what firms can
do—thereby targeting the tools firms use to
exploit consumers—can raise social welfare.
For instance, Heidhues and Kőszegi (2010)
show that because the large welfare losses partially naïve consumers suffer in credit markets
are driven by large fees, prohibiting large fees
for small deviations from contract terms often
raises welfare for any combination of naïve
and sophisticated consumers in the population. Similarly, Inderst and Ottaviani (2012)
and Murooka (2013) show that because the
distorted advice financial advisors provide is
driven by the large commissions firms pay,
capping or banning commissions, or requiring
commissions to be uniform across products,
can raise consumer welfare.30
30 Korobkin (2003) provides an interesting legal perspective on exploitative consumer contracts. Similarly to
the economic research in this section, he argues that when
consumers do not understand all aspects of the contract,
sellers have an incentive to include inefficient terms that
favor themselves. As a remedy, he proposes a modification of the unconscionability doctrine, a legal doctrine that
renders contract features a party had no effective choice
over invalid. Although this is not the prevailing legal interpretation, Korobkin argues that a consumer who did not
understand part of the contract could not have had effective choice over it. And in a completely different approach,
Bubb and Kaufman (2011) propose that different ownership structures, such as customer-owned firms, can lower
firms’ incentive to exploit consumer mistakes.
Problems with government intervention. A
number of authors point out potential problems with market intervention. Grubb (2012)
considers services, such as mobile phones
and bank overdraft protection, for which
consumers may not know the marginal price
of the next unit of service, and asks whether
requiring firms to disclose this information
at the point of sale increases welfare. If consumers correctly anticipate their probability
of running into penalties, such regulation
can actually hurt because it interferes with
efficient screening by firms. Intuitively, penalty fees for high usage prevent high-value
consumers from taking the contracts offered
to low-value consumers; yet because consumers do not know when they apply, these
fees do not distort the consumption of lowvalue consumers.
Kosfeld and Schüwer (2011) establish
that, since the inefficiency in Gabaix and
Laibson’s 2006 model of shrouded attributes
is due to sophisticated consumers’ efforts to
avoid the add-on, in such a setting, education—modeled as turning a portion of naïve
consumers into sophisticated consumers—
can lower social welfare. And Warren and
Wood (2010) show that when naïve consumers misperceive add-on prices, there is no
budget-balanced regulation that consumers
will perceive as improving their welfare, so
that there is no welfare-improving regulation
that citizens will vote for. Intuitively, a competitive market redistributes income from
naive to sophisticated consumers, and since
no consumer believes that she is naïve, she
believes she benefits from the redistribution.
6.3 Potential Future Directions
While researchers have identified a number of features of exploitative contracts,
overall this literature seems to be in its
infancy, with a variety of open questions. I
highlight some major issues. First, while
many researchers employ a static framework
or assume that firms and consumers sign
Kőszegi: Behavioral Contract Theory
exclusive contracts, it seems important to
understand the effect of ex post competition
and the possibility of switching for contract
structure. Gottlieb (2008) takes a first step in
this research agenda, exploring the dynamics of contracting with present-biased agents
in the setting of DellaVigna and Malmendier
(2004). He shows that ex post competition
leaves contracts for products with immediate costs and future benefits unaffected, but
renders contracts for products with immediate benefits and future costs equivalent to
spot markets. As an example in the context
of smoking, while a retailer and a presentbiased consumer might prefer to sign an
exclusive contract in which (to restrain her
future smoking) the consumer is paid a lump
sum ex ante and buys overpriced cigarettes
from the retailer ex post, this contract is
ineffective if the consumer can go to other
retailers. Nevertheless, in some markets for
products with immediate benefits, such as
credit cards, consumers do not switch easily
to competitors, and firms seem to take advantage of this (Ausubel 1991). Procrastination
on switching offers a plausible explanation,
but this explanation and especially its implications for equilibrium contracts have not
been formally analyzed.
A noticeable aspect of existing exploitative
contracts seems to be that exploitative features are often masked as having alternative
uses. For instance, while the primary purpose of a high credit-card late fee is likely to
exploit a consumer’s inattention to this fee or
her optimism that she will not pay late, one
can make the argument that the purpose is to
overcome the moral-hazard problem that she
might delay repayment. An interesting question for behavioral contract theory is why
exploitative features that can in no plausible
environment be useful are so rare. Some
form of metasophistication on the part of
consumers—whereby consumers might realize the possibility of exploitation, and pure
exploitative features might seem suspicious
1109
to them—could be involved. Alternatively,
this pattern may be due to regulation or the
threat of regulation, making it more difficult
to conclusively establish that the feature in
question is deceptive.
Another poorly understood issue is how to
improve consumer decision making, including how to educate or inform consumers
about their own behavior. The likely impact
of proposed choice engines to improve consumer decision making, and their optimal
regulation, also deserves closer theoretical
scrutiny. The concern arises that the platform
for exploitation shifts from firms to choice
engines—which become like other intermediaries who deceive consumers—and it
is unclear what business model for choice
engines will prevent this. In particular, if
choice engines are paid for finding a match
between buyers and sellers, then to attract
consumers, they appear to have similar
incentives to mislead consumers as firms do.
Furthermore, because choice engines can
analyze offers much more quickly, their presence can lead to a proliferation of products,
including products that take advantage of
very specific or rare mistakes by consumers.
Finally, researchers have studied contracting to exploit agents who are naïve about
their suboptimal behavior much more than
they have studied contracting to help agents
who are sophisticated about their suboptimal
behavior (e.g., commitment contracts)—
although such sophisticated agents should
in principle demand help. Indeed, this
asymmetry in the literature seems to reflect
an asymmetry in the real-life contracts we
observe. Whether there are more exploitative than “helping” contracts, and, if so,
why, is an important area for future research.
There are at least two main possibilities.
First, the market mechanism might, for
some reason, favor the exploitation of biases
rather than their mitigation. Second, it might
be the case that most individuals are not sufficiently sophisticated to demand help, or are
1110
Journal of Economic Literature, Vol. LII (December 2014)
averse to helping contracts for another reason, such as being averse to committing their
own behavior.
7. Other Topics
This section covers two further small literatures in behavioral contract theory.
7.1 Incomplete Contracts
A contract is incomplete if it leaves the
details of some transaction that the parties
care about to be determined at a later time.
In the classical literature on incomplete contracts, such as Grossman and Hart (1986)
and Hart and Moore (1990), parties make
interim investments into the relationship,
and because the incomplete contract cannot
provide direct incentives for these investments, they are typically inefficient. The
main contracting problem is to maximize the
efficiency of the interim decisions, primarily
through the allocation of ownership and control rights.
The psychology-and-economics literature
on incomplete contracts is relatively small.
Papers in this literature study ways in which
individuals’ reactions to a contract generate
inefficiencies ex post (in contrast to the classical literature, where inefficiencies arise at
the investment stage) and how this affects
optimal contracts and the optimal allocation
of ownership rights.
In the most influential contribution of this
literature, Hart and Moore (2008) analyze a
buyer–seller relationship when the contract
parties write initially cannot condition on the
state of the world that determines the optimal terms of trade, and at the ex post stage
an aggrieved party can inefficiently “shade”
on her performance in a way that is impossible to prevent contractually. As a simple
example, while a court may be able to verify
that a professor delivered an agreed-to lecture, it cannot verify that the lecture was
good, and hence a professor can shade by
elivering a bad lecture. Hart and Moore
d
assume that a party is aggrieved and shades
if she gets a worse outcome than the best
possible under the initial contract.31 The
way to avoid aggrieving a party and shading,
therefore, is to fix or restrict the terms in the
ex ante contract; for instance, the parties
may agree to trading at a fixed price. This will
result in a failure to trade in circumstances
when the price does not fall between the
buyer’s value and the seller’s cost, but economizes on shading costs, and the optimal
contract trades off these two considerations.
Furthermore, among multiple terms the parties eventually decide on, it is most important
to fix the ones that create the most conflict of
interest—and therefore the most potential
for shading. Indeed, since the parties have
a direct conflict regarding the price, an optimal contract might fix the price but not the
full description of the good to be delivered.
Finally, to minimize shading, the party who
cares more about the good should have the
right to specify it ex post. In the extreme, if
the seller is almost indifferent between different products but the buyer cares a lot—
such as in a firm where the employee might
perform a number of similarly difficult tasks,
only a few of which are useful—it is optimal
for the buyer to have control rights. If the
opposite is the case—such as when the seller
is subject to large cost shocks—the seller
should be in control.32
Hart (2009) applies a similar model to a
holdup problem, where holding up the other
party results in inefficient shading in the
31 Psychologically, this means that a party is not
aggrieved if the terms of the initial contract are bad, but is
aggrieved if the terms of the renegotiated transaction are
bad (in her view). Hart and Moore (2008) explain this distinction by arguing that there is a change in circumstances:
the final agreement often occurs in situations of bilateral
monopoly, whereas the initial contract is agreed to in a
more competitive, “objective” setting.
32 See also Hart and Holmstrom (2010) for an application of this approach to firm scope—whether and when
units should integrate to a single firm or merely cooperate.
Kőszegi: Behavioral Contract Theory
renegotiated contract. A buyer and a seller
agree to a fixed price ex ante and, when the
state of the world is realized, decide whether
to hold up the other party by refusing to
trade. Because holdup results in shading, a
party only chooses it if the resulting price is
sufficiently better than the price agreed to in
the contract. Hart shows that an appropriate
allocation of asset ownership can lower the
probability of inefficient holdup. In particular, it is optimal to assign asset ownership to
the party who faces more uncertainty regarding her value from trade; since this increases
the party’s outside option exactly when her
value from trade is high, it makes holding her
up less profitable. This implies that—unlike
in the classical model, where asset ownership
is determined by the importance of interim
investments—asset ownership is determined
by the importance of uncertainty in a party’s
value from trade.
Herweg and Schmidt (2012) take a somewhat different approach to how contracts
affect preferences over trades. They suppose
that a loss-averse buyer and a loss-averse
seller write a contract specifying the nature
and price of the good to be delivered at a
later date, and the contract serves as a reference point for outcomes ex post. Since a
party evaluates a better-than-contracted
term (e.g., an increase in price for the seller)
as a gain and a worse-than-contracted term
as a loss, and loss aversion implies that she is
more sensitive to the loss, parties are reluctant to compromise and trade away from the
contract. In particular, if the ex post optimal terms are close to the specified terms,
the parties stick to the contract, and even if
the ex post optimal terms are far, the parties adjust the terms only partially, in either
case generating (material) inefficiency.
Here, inefficiency arises when there is a specific contract, whereas in Hart and Moore
(2008), inefficiency arises when there is only
a vague contract. An immediate implication
is that the parties might prefer not to write
1111
a contract so as not to set a reference point,
or—to optimize the insufficient adjustment
ex post—might prefer to write a kind of
“compromise” contract that they know they
will trade away from. The paper also derives
two less immediate implications. First, it is
optimal to write a specific contract rather
than rely on ownership rights to incentivize
relationship-specific investments if there is
little uncertainty or parties are not very loss
averse. Second, an employment contract
dominates a fixed performance contract if
the scope for inefficient abuse is small relative to the renegotiation costs generated by
loss aversion due to uncertainty.
7.2 Environment Design
A few papers investigate, under various
psychologically motivated assumptions, how
to design individuals’ decision-making environment to achieve specific goals.
Optimal nudges. Psychology and economics has had a major practical impact in motivating “soft paternalism”—designing policies
such that individuals with a tendency to
behave suboptimally make better decisions,
but others are either free to choose as they
would otherwise (Thaler and Sunstein 2003;
Thaler and Sunstein 2008), or are not hurt
much (Camerer et al., 2003). Some work
can be thought of as providing theoretical
guidance as to how such “nudges” should be
designed.
O’Donoghue and Rabin (1999b) analyze the problem of a principal employing
a naïve present-biased agent to complete a
single task with an uncertain cost of effort,
where he can pay the agent depending on
when she completes the task. The principal
faces a delay cost, yet because in any given
period the effort cost might be high, efficiency requires that the agent sometimes
wait. Consistent with the soft paternalist
agenda but quite distinct from the literature on exploitative contracts, O’Donoghue
1112
Journal of Economic Literature, Vol. LII (December 2014)
and Rabin assume that the principal’s goal
is to choose the most efficient (rather than
profit-maximizing) incentive scheme. Note
that with time-consistent preferences, such
an incentive scheme is obvious: the principal
imposes a punishment for delay that is exactly
equal to his delay cost, generating a first-best
outcome. When the agent is present-biased
and the principal does not know the task-cost
distribution, in contrast, the first-best outcome is not achievable. Intuitively, since for
high average costs the agent is more prone
to procrastinate, the principal needs severe
punishment for delay to induce her to complete the task efficiently; but if he imposes
such a punishment, an agent with low average costs will tend to complete the task too
soon. The optimal contract, then, resembles
a deadline: punishment for delay is relatively
mild for a while to allow an agent with low
average cost to delay if necessary, but severe
after a deadline to discourage an agent with
high average costs from procrastinating.
Carroll et al. (2009) study optimal defaults
for retirement savings decisions when
employees have time-inconsistent tastes for
immediate gratification, and have heterogeneous optimal savings rates. Employees are
initially assigned the default savings rate,
and in each period draw a stochastic cost at
which they can change their savings rate to
the optimal one, suffering a flow utility loss
if they do not. Employees’ problem is that
(due to their time inconsistency) they may
procrastinate paying the cost. Carroll et al.
establish that if employees have a large timeinconsistency problem and are not too heterogeneous, it is optimal to set the default
at the optimal rate for the population distribution. If employees are very heterogenous
or the time inconsistency is not so serious, in
contrast, the optimal policy involves “active
decisions”: the default is set at such an unattractive level that all employees are forced to
pay the cost immediately. And in in-between
cases, a compromise policy is optimal: the
default is set such that it is good enough
for part of the population, while the rest of
the population is forced to pay the cost and
switch to the optimal savings rate.
Information optimization with emotions.
Caplin and Eliaz (2003) and Schweizer and
Szech (2013) characterize optimal medical
tests when individuals derive anticipatory
utility from their beliefs about their health
status, and find bad news more aversive
than good news pleasant. In both papers,
the principal can commit to an informationrevelation policy that is conditional on an
individual’s status. Caplin and Eliaz establish
that an optimal policy for stopping the spread
of HIV gives only noisy bad news: it certifies
some, but not all, individuals who are uninfected, and gives the same signal to everyone else. This policy protects individuals
from receiving very bad news, yet provides
sufficient information to help match uninfected individuals. Using methods similar
to those in Kamenica and Gentzkow (2011),
Schweizer and Szech (2013) show that such
a policy also emerges as the optimal way to
reveal information about infection status to
an individual who uses the information to
make better decisions.
7.3 Potential Future Directions
Most of the literature on incomplete contracts is based on theories of how a contract
agreed to by the parties affects future preferences over outcomes. While this approach
has produced important insights, it seems
both too broad and too narrow for incomplete
contracts. It is too broad because the question of how contracts affect preferences is an
important topic for research quite independently of whether the contract is incomplete.
Identifying the psychological underpinnings
of how contracts affect preferences—including the extent to which these can be derived
from psychological phenomena, such as the
role of expectations or social preferences, that
Kőszegi: Behavioral Contract Theory
go beyond contracts, and the extent to which
they are specific to contracts—seems to be
an important agenda for future research. At
the same time, the question of how contracts
affect preferences is also too narrow for
incomplete contracts because psychological
phenomena that are not directly about this
question are also likely to have implications
for incomplete contracts.
A long-recognized issue at the foundations
of incomplete contracts is whether and when
it is reasonable to assume that contracts will
be incomplete. Since rational agents can
often write optimal contracts that render
standard arguments for incomplete contracts
irrelevant (for example Moore 1992; Maskin
and Tirole 1999), some researchers have
conjectured that ideas from psychology and
economics might help provide more compelling foundations for the incompleteness of
contracts. This is still an important problem
for future research.
Relative to the attention that has been paid
to the question of soft paternalism at least
since the success of Thaler and Benartzi’s
(2004) Save More Tomorrow plan, the dearth
of theoretical research on such policies is
striking. One potential reason is that most
real-life nudges take advantage of the default
effect—that individuals tend to stick with an
option they have chosen or to which they have
been automatically assigned—and we do not
have good theories for why defaults influence
behavior in such a powerful way.33 Yet providing much better theoretical guidance on soft
paternalism seems essential for the continued
practical success of the agenda, especially in
figuring out what kinds of policies are likely
to work and which policies push people in the
right direction. Theoretical work is also necessary for delineating the limits of the nudge
33 Most likely, there are multiple psychological reasons
for people not to switch away from defaults, and the effect
is robust and powerful because in most situations multiple
forces operate at the same time.
1113
agenda—that is, when more heavy-handed
policies are desirable.
In research on emotions and information in contract theory, authors have largely
confined themselves to Bayesian models,
although anticipatory utility might lead to
non-Bayesian information processing, and
beliefs formed in a non-Bayesian way can
change emotions. Oster, Shoulson, and
Dorsey (2013), for instance, argue that facts
regarding testing and economic behavior
among patients at risk for Huntington’s disease are only consistent with n
on-Bayesian
models of anticipatory utility, such as
Brunnermeier and Parker (2005). Thinking
about how to optimize information revelation with non-Bayesian beliefs seems like an
interesting agenda.
8. Conclusion
This section discusses a few general issues
regarding the state and direction of the literature on behavioral contract theory, and
applied behavioral-economics theory more
generally. When applying a behavioral-economics theory to an economic setting, the
all-too common tendency of much research
is to study how the parameters of the behavioral theory affect predictions. In studying
the implications of present bias for credit,
for instance, a researcher might analyze how
less and more present-biased individuals
behave. While such analyses are useful for
understanding the model and comparing it
to a classical one, ultimately economists are
more interested in comparative statics with
respect to variables typically studied in economic analysis. To continue with the previous example, a researcher could study how
the same present-biased individual responds
to different kinds of credit or to changes in
the interest rate. Such predictions are both
economically more relevant and easier to
test than those deriving from manipulating
the behavioral model itself.
1114
Journal of Economic Literature, Vol. LII (December 2014)
A central constraint on progress in behavioral economics is the scarcity of compelling
portable models of individual decision making that researchers can apply to economic
settings. Much of behavioral contract theory
uses only the four models introduced in section 2, highlighting both that useful models
can be taken up by many researchers, and
that there are too few portable models. As
one particular example, developing portable
psychologically based models of limited or
distorted attention is a first-order question.
References
Akerlof, George A. 1982. “Labor Contracts as Partial
Gift Exchange.” Quarterly Journal of Economics 97
(4): 543–69.
Akerlof, George A., and Janet L. Yellen. 1988. “Fairness
and Unemployment.” American Economic Review 78
(2): 44–49.
Akerlof, George A., and Janet L. Yellen. 1990. “The
Fair Wage-Effort Hypothesis and Unemployment.”
Quarterly Journal of Economics 105 (2): 255–83.
Alger, Ingela, and Régis Renault. 2006. “Screening
Ethics When Honest Agents Care about Fairness.”
International Economic Review 47 (1): 59–85.
Amador, Manuel, Iván Werning, and George-Marios
Angeletos. 2006. “Commitment vs. Flexibility.”
Econometrica 74 (2): 365–96.
Armstrong, Mark, and John Vickers. 2012. “Consumer
Protection and Contingent Charges.” Journal of Economic Literature 50 (2): 477–93.
Armstrong, Mark, and Jidong Zhou. 2011. “Paying for
Prominence.” Economic Journal 121 (556): F368–95.
Augenblick, Ned. 2011. “Consumer and Producer
Behavior in the Market for Penny Auctions: A Theoretical and Empirical Analysis.” Unpublished.
Ausubel, Lawrence M. 1991. “The Failure of Competition in the Credit Card Market.” American Economic
Review 81 (1): 50–81.
Bar-Gill, Oren. 2006. “Bundling and Consumer
Misperception.” University of Chicago Law Review
73 (1): 33–61.
Bar-Gill, Oren, and Franco Ferrari. 2010. “Informing Consumers about Themselves.” Erasmus Law
Review 3 (2): 93–119.
Bartling, Björn. 2011. “Relative Performance or Team
Evaluation? Optimal Contracts for Other-Regarding
Agents.” Journal of Economic Behavior and Organization 79 (3): 183–93.
Bartling, Björn, and Ferdinand A. von Siemens. 2010.
“The Intensity of Incentives in Firms and Markets:
Moral Hazard with Envious Agents.” Labour Economics 17 (3): 598–607.
Bénabou, Roland, and Jean Tirole. 2003. “Intrinsic and
Extrinsic Motivation.” Review of Economic Studies
70 (3): 489–520.
Bénabou, Roland, and Jean Tirole. 2006. “Incentives
and Prosocial Behavior.” American Economic Review
96 (5): 1652–78.
Bierbrauer, Felix, and Nick Netzer. 2012. “Mechanism Design and Intentions.” University of Cologne
Department of Economics Working Paper 53.
Bolton, Patrick, and Mathias Dewatripont. 2005. Contract Theory. Cambridge, Mass. and London: MIT
Press.
Bolton, Patrick, and Antoine Faure-Grimaud. 2010.
“Satisficing Contracts.” Review of Economic Studies
77 (3): 937–71.
Brunnermeier, Markus K., and Jonathan A. Parker.
2005. “Optimal Expectations.” American Economic
Review 95 (4): 1092–1118.
Bubb, Ryan, and Alex Kaufman. 2011. “Consumer
Biases and Mutual Ownership.” New York University
School of Law Law and Economics Research Paper
11-35.
Camerer, Colin F., Teck-Hua Ho, and Juin-Kuan
Chong. 2004. “A Cognitive Hierarchy Model of
Games.” Quarterly Journal of Economics 119 (3):
861–98.
Camerer, Colin F., Samuel Issacharoff, George Loewenstein, Ted O’Donoghue, and Matthew Rabin.
2003. “Regulation for Conservatives: Behavioral
Economics and the Case for ‘Asymmetric Paternalism’.” University of Pennsylvania Law Review 151
(3): 1211–54.
Caplin, Andrew, and Kfir Eliaz. 2003. “AIDS Policy
and Psychology: A Mechanism-Design Approach.”
RAND Journal of Economics 34 (4): 631–46.
Carbajal, Juan Carlos, and Jeffrey C. Ely. 2012.
“Optimal Contracts for Loss Averse Consumers.”
Unpublished.
Carmichael, Lorne, and W. Bentley MacLeod. 2003.
“Caring About Sunk Costs: A Behavioral Solution
to Holdup Problems with Small Stakes.” Journal of
Law, Economics and Organization 19 (1): 106–18.
Carroll, Gabriel D., James J. Choi, David Laibson,
Brigitte C. Madrian, and Andrew Metrick. 2009.
“Optimal Defaults and Active Decisions.” Quarterly
Journal of Economics 124 (4): 1639–74.
Chiappori, Pierre-Andre, and Bernard Salanie. 2000.
“Testing for Asymmetric Information in Insurance
Markets.” Journal of Political Economy 108 (1):
56–78.
Cox, James C., Vernon L. Smith, and James M. Walker.
1988. “Theory and Individual Behavior of First-Price
Auctions.” Journal of Risk and Uncertainty 1 (1):
61–99.
Cramton, Peter, Emel Filiz-Ozbay, Erkut Y. Ozbay, and
Pacharasut Sujarittanonta. 2012. “Fear of Losing in
a Clock Auction.” Review of Economic Design 16
(2–3): 119–34.
Crawford, Vincent P., and Nagore Iriberri. 2007.
“Level-k Auctions: Can a Nonequilibrium Model of
Strategic Thinking Explain the Winner’s Curse and
Overbidding in Private-Value Auctions?” Econometrica 75 (6): 1721–70.
Kőszegi: Behavioral Contract Theory
Crawford, Vincent P., Tamar Kugler, Zvika Neeman,
and Ady Pauzner. 2009. “Behaviorally Optimal Auction Design: Examples and Observations.” Journal of
the European Economic Association 7 (2–3): 377–87.
Daido, Kohei, and Takeshi Murooka. 2012. “Team
Incentives and Reference-Dependent Preferences.”
http://papers.ssrn.com/sol3/papers.
cfm?abstract_id=1922366.
Deci, Edward L. 1975. Intrinsic Motivation. New York:
Plenum Publishing Company.
Deci, Edward L., Richard Koestner, and Richard M.
Ryan. 1999. “A Meta-analytic Review of Experiments Examining the Effects of Extrinsic Rewards
on Intrinsic Motivation.” Psychological Bulletin 125
(6): 627–68.
de la Rosa, Leonidas Enrique. 2011. “Overconfidence
and Moral Hazard.” Games and Economic Behavior
73 (2): 429–51.
DellaVigna, Stefano. 2009. “Psychology and Economics: Evidence from the Field.” Journal of Economic
Literature 47 (2): 315–72.
DellaVigna, Stefano, and Ulrike Malmendier. 2004.
“Contract Design and Self-Control: Theory and
Evidence.” Quarterly Journal of Economics 119 (2):
353–402.
de Meza, David, and David C. Webb. 2007. “Incentive
Design under Loss Aversion.” Journal of the European Economic Association 5 (1): 66–92.
Desiraju, Ramarao, and David E. M. Sappington. 2007.
“Equity and Adverse Selection.” Journal of Economics and Management Strategy 16 (2): 285–318.
Eisenhuth, Roland. 2010. “Auction Design with Loss
Averse Bidders: The Optimality of All Pay Mechanisms.” Munich Personal RePEc Archive Paper
23357.
Eisenhuth, Roland. 2013. “Reference Dependent
Mechanism Design.” Unpublished.
Eisenhuth, Roland, and Mara Ewers. 2012. “Auctions
with Loss Averse Bidders.” Unpublished.
Eliaz, Kfir, and Ran Spiegler. 2006. “Contracting with
Diversely Naïve Agents.” Review of Economic Studies 73 (3): 689–714.
Eliaz, Kfir, and Ran Spiegler. 2008. “Consumer Optimism and Price Discrimination.” Theoretical Economics 3 (4): 459–97.
Eliaz, Kfir, and Ran Spiegler. 2014. “Reference-Dependence and Labor-Market Fluctuations.” In NBER
Macroeconomics Annual 2013, edited by Jonathan
Parker and Michael Woodford, 159–200. Chicago
and London: University of Chicago Press.
Ellingsen, Tore, and Magnus Johannesson. 2008. “Pride
and Prejudice: The Human Side of Incentive Theory.” American Economic Review 98 (3): 990–1008.
Engelbrecht-Wiggans, Richard. 1989. “The Effect of
Regret on Optimal Bidding in Auctions.” Management Science 35 (6): 685–92.
Englmaier, Florian, and Stephen Leider. 2012. “Contractual and Organizational Structure with Reciprocal
Agents.” American Economic Journal: Microeconomics 4 (2): 146–83.
Englmaier, Florian, and Achim Wambach. 2010. “Opti-
1115
mal Incentive Contracts under Inequity Aversion.”
Games and Economic Behavior 69 (2): 312–28.
Esteban, Susanna, and Eiichi Miyagawa. 2006. “Temptation, Self-Control, and Competitive Nonlinear
Pricing.” Economics Letters 90 (3): 348–55.
Esteban, Susanna, Eiichi Miyagawa, and Matthew
Shum. 2007. “Nonlinear Pricing with Self-Control
Preferences.” Journal of Economic Theory 135 (1):
306–38.
Eyster, Erik, and Matthew Rabin. 2005. “Cursed Equilibrium.” Econometrica 73 (5): 1623–72.
Falk, Armin, and Michael Kosfeld. 2006. “The Hidden
Costs of Control.” American Economic Review 96
(5): 1611–30.
Fang, Hanming, and Giuseppe Moscarini. 2005.
“Morale Hazard.” Journal of Monetary Economics 52
(4): 749–77.
Farrell, Joseph, and Paul Klemperer. 2007. “Coordination and Lock-In: Competition with Switching Costs
and Network Effects.” In Handbook of Industrial
Organization, Volume 3, edited by Mark Armstrong
and Robert H. Porter, 1967–2072. Amsterdam: Elsevier, North-Holland.
Fehr, Ernst, Lorenz Goette, and Christian Zehnder.
2009. “A Behavioral Account of the Labor Market:
The Role of Fairness Concerns.” Annual Review of
Economics 1: 355–84.
Fehr, Ernst, Alexander Klein, and Klaus M. Schmidt.
2007. “Fairness and Contract Design.” Econometrica
75 (1): 121–54.
Fehr, Ernst, and Klaus M. Schmidt. 1999. “A Theory of
Fairness, Competition, and Cooperation.” Quarterly
Journal of Economics 114 (3): 817–68.
Fehr, Ernst, and Klaus M. Schmidt. 2004. “Fairness and
Incentives in a Multi-task Principal–Agent Model.”
Scandinavian Journal of Economics 106 (3): 453–74.
Filiz-Ozbay, Emel, and Erkut Y. Ozbay. 2007. “Auctions with Anticipated Regret: Theory and Experiment.” American Economic Review 97 (4): 1407–18.
Finkelstein, Amy, and Kathleen McGarry. 2006. “Multiple Dimensions of Private Information: Evidence
from the Long-Term Care Insurance Market.” American Economic Review 96 (4): 938–58.
Gabaix, Xavier, and David Laibson. 2006. “Shrouded
Attributes, Consumer Myopia, and Information Suppression in Competitive Markets.” Quarterly Journal
of Economics 121 (2): 505–40.
Galperti, Simone. 2012. “Commitment, Flexibility,
and Optimal Screening of Time Inconsistency.”
Unpublished.Garicano, Luis, and Andrea Prat. 2013.
“Organizational Economics with Cognitive Costs.”
In Advances in Economics and Econometrics: Tenth
World Congress, Volume I, Economic Theory, edited
by Daron Acemoglu, Manuel Arellano, and Eddie
Dekel, 342–88. Cambridge and New York: Cambridge University Press.
Gervais, Simon, J. B. Heaton, and Terrance Odean.
2011. “Overconfidence, Compensation Contracts,
and Capital Budgeting.” Journal of Finance 66 (5):
1735–77.
Gneezy, Uri, and Aldo Rustichini. 2000. “A Fine Is a
1116
Journal of Economic Literature, Vol. LII (December 2014)
Price.” Journal of Legal Studies 29 (1): 1–17.
Gottlieb, Daniel. 2008. “Competition over TimeInconsistent Consumers.” Journal of Public Economic Theory 10 (4): 673–84.
Gottlieb, Daniel, and Kent Smetters. 2012. “Narrow
Framing and Life Insurance.” National Bureau of
Economic Research Working Paper 18601.
Grossman, Sanford J., and Oliver D. Hart. 1986. “The
Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration.” Journal of Political
Economy 94 (4): 691–719.
Grubb, Michael D. 2009. “Selling to Overconfident
Consumers.” American Economic Review 99 (5):
1770–1807.
Grubb, Michael D. 2012. “Consumer Inattention and
Bill-Shock Regulation.” Massachusetts Institute of
Technology Sloan School Working Paper 4987-12.
Gul, Faruk, and Wolfgang Pesendorfer. 2001. “Temptation and Self-Control.” Econometrica 69 (6):
1403–35.
Hahn, Jong-Hee, Jinwoo Kim, Sang-Hyun Kim,
and Jihong Lee. 2010. “Screening Loss Averse
Consumers.”
http://papers.ssrn.com/sol3/papers.
cfm?abstract_id=1678825.
Harrison, J. Michael, and David M. Kreps. 1978.
“Speculative Investor Behavior in a Stock Market
with Heterogeneous Expectations.” Quarterly Journal of Economics 92 (2): 323–36.
Hart, Oliver. 2009. “Hold-Up, Asset Ownership, and
Reference Points.” Quarterly Journal of Economics
124 (1): 267–300.
Hart, Oliver, and Bengt Holmstrom. 2010. “A Theory
of Firm Scope.” Quarterly Journal of Economics 125
(2): 483–513.
Hart, Oliver, and John Moore. 1990. “Property Rights
and the Nature of the Firm.” Journal of Political
Economy 98 (6): 1119–58.
Hart, Oliver, and John Moore. 2008. “Contracts as Reference Points.” Quarterly Journal of Economics 123
(1): 1–48.
Heidhues, Paul, and Botond Kőszegi. 2008. “Competition and Price Variation When Consumers Are
Loss Averse.” American Economic Review 98 (4):
1245–68.
Heidhues, Paul, and Botond Kőszegi. 2010. “Exploiting
Naivete about Self-Control in the Credit Market.”
American Economic Review 100 (5): 2279–303.
Heidhues, Paul, and Botond Kőszegi. 2013. “Seller
Information about Consumer Naivete Lowers Welfare.” Unpublished.
Heidhues, Paul, Botond Kőszegi, and Takeshi Murooka. 2012. “Inferior Products and Profitable Deception.” Unpublished.
Herold, Florian. 2010. “Contractual Incompleteness as
a Signal of Trust.” Games and Economic Behavior 68
(1): 180–91.
Herweg, Fabian, and Klaus M. Schmidt. 2012. “A
Theory of Ex Post Inefficient Renegotiation.”
Unpublished.
Herweg, Fabian, and Konrad Mierendorff. 2013.
“Uncertain Demand, Consumer Loss Aversion, and
Flat-Rate Tariffs.” Journal of the European Economic
Association 11 (2): 399–432.
Herweg, Fabian, Daniel Müller, and Philipp Weinschenk. 2010. “Binary Payment Schemes: Moral
Hazard and Loss Aversion.” American Economic
Review 100 (5): 2451–77.
Holmstrom, Bengt. 1979. “Moral Hazard and Observability.” Bell Journal of Economics 10 (1): 74–91.
Holmstrom, Bengt, and Paul Milgrom. 1991. “Multitask Principal–Agent Analyses: Incentive Contracts,
Asset Ownership, and Job Design.” Journal of Law,
Economics, and Organization 7 (Special Issue):
24–52.
Immordino, Giovanni, Anna Maria C. Menichini, and
Maria Grazia Romano. 2011. “A Simple Impossibility Result in Behavioral Contract Theory.” Economics
Letters 113 (3): 307–09.
Inderst, Roman, and Marco Ottaviani. 2012. “Financial Advice.” Journal of Economic Literature 50 (2):
494–512.
Itoh, Hideshi. 2004. “Moral Hazard and Other-Regarding Preferences.” Japanese Economic Review 55 (1):
18–45.
Jehiel, Philippe. 2005. “Analogy-Based Expectation
Equilibrium.” Journal of Economic Theory 123 (2):
81–104.
Jehiel, Philippe, and Laurent Lamy. 2012. “Absolute
Auctions and Secret Reserve Prices: Why Are They
Used?” Unpublished.
Jofre, Alejandro, Sofia Moroni, and Andrea Repetto.
2012. “Dynamic Contracts under Loss Aversion.”
Unpublished.
Kahneman, Daniel, and Amos Tversky. 1979. “Prospect
Theory: An Analysis of Decision under Risk.” Econometrica 47 (2): 263–92.
Kamenica, Emir. 2012. “Behavioral Economics and
Psychology of Incentives.” Annual Review of Economics 4: 427–52.
Kamenica, Emir, and Matthew Gentzkow. 2011.
“Bayesian Persuasion.” American Economic Review
101 (6): 2590–615.
Kamenica, Emir, Sendhil Mullainathan, and Richard
Thaler. 2011. “Helping Consumers Know Themselves.” American Economic Review 101 (3): 417–22.
Korobkin, Russell. 2003. “Bounded Rationality,
Standard Form Contracts, and Unconscionability.”
University of Chicago Law Review 70 (4): 1203–95.
Kosfeld, Michael, and Ulrich Schüwer. 2011. “Add-on
Pricing, Naïve Consumers, and the Hidden Welfare
Costs of Education.” Centre for Economic Policy
Research Discussion Paper 8636.
Kosfeld, Michael, and Ferdinand A. von Siemens. 2011.
“Competition, Cooperation, and Corporate Culture.”
RAND Journal of Economics 42 (1): 23–43.
Kőszegi, Botond, and Matthew Rabin. 2006. “A Model
of Reference-Dependent Preferences.” Quarterly
Journal of Economics 121 (4): 1133–65.
Kőszegi, Botond, and Matthew Rabin. 2007. “Reference-Dependent Risk Attitudes.” American Economic Review 97 (4): 1047–73.
Kőszegi, Botond, and Matthew Rabin. 2009. “Refer-
Kőszegi: Behavioral Contract Theory
ence-Dependent Consumption Plans.” American
Economic Review 99 (3): 909–36.
Kucuksenel, Serkan. 2012. “Behavioral Mechanism
Design.” Journal of Public Economic Theory 14 (5):
767–89.
Laibson, David. 1997. “Golden Eggs and Hyperbolic
Discounting.” Quarterly Journal of Economics 112
(2): 443–78.
Lal, Rajiv, and Carmen Matutes. 1994. “Retail Pricing
and Advertising Strategies.” Journal of Business 67
(3): 345–70.
Lange, Andreas, and Anmol Ratan. 2010. “Multidimensional Reference-Dependent Preferences
in Sealed-Bid Auctions—How (Most) Laboratory
Experiments Differ From the Field.” Games and
Economic Behavior 68 (2): 634–45.
List, John A. 2007. “On the Interpretation of Giving in
Dictator Games.” Journal of Political Economy 115
(3): 482–93.
Macera, Rosario. 2012. “Intertemporal Incentives
under Loss Aversion.” Unpublished.
Malmendier, Ulrike, and Young Han Lee. 2011. “The
Bidder’s Curse.” American Economic Review 101 (2):
749–87.
Manove, Michael, and A. Jorge Padilla. 1999. “Banking
(Conservatively) with Optimists.” RAND Journal of
Economics 30 (2): 324–50.
Maskin, Eric, and Jean Tirole. 1999. “Unforeseen Contingencies and Incomplete Contracts.” Review of
Economic Studies 66 (1): 83–114.
Matsushima, Hitoshi. 2008. “Role of Honesty in Full
Implementation.” Journal of Economic Theory 139
(1): 353–59.
Milgrom, Paul R., and Robert J. Weber. 1982. “A Theory of Auctions and Competitive Bidding.” Econometrica 50 (5): 1089–1122.
Moore, John. 1992. “Implementation in Environments with Complete Information.” In Advances in
Economic Theory, edited by Jean-Jacques Laffont,
181–282. Cambridge and New York: Cambridge
University Press.
Morgan, John, Ken Steiglitz, and George Reis. 2003.
“The Spite Motive and Equilibrium Behavior in Auctions.” B. E. Journal of Economic Analysis and Policy
2 (1): 1–27.
Mullainathan, Sendhil, Joshua Schwartzstein, and William J. Congdon. 2012. “A Reduced-Form Approach
to Behavioral Public Finance.” Annual Review of
Economics 4: 511–40.
Murooka, Takeshi. 2013. “Deception under Competitive Intermediation.” Unpublished.
Netzer, Nick, and Armin Schmutzler. 2012. “Explaining Gift-Exchange—The Limits of Good Intentions.”
Unpublished.O’Donoghue, Ted, and Matthew
Rabin. 1999a. “Doing It Now or Later.” American
Economic Review 89 (1): 103–24.
O’Donoghue, Ted, and Matthew Rabin. 1999b. “Incentives for Procrastinators.” Quarterly Journal of Economics 114 (3): 769–816.
O’Donoghue, Ted, and Matthew Rabin. 2001. “Choice
and Procrastination.” Quarterly Journal of Econom-
1117
ics 116 (1): 121–60.
O’Donoghue, Ted, and Matthew Rabin. 2003. “Studying Optimal Paternalism, Illustrated by a Model
of Sin Taxes.” American Economic Review 93 (2):
186–91.
Oster, Emily, Ira Shoulson, and E. Ray Dorsey. 2013.
“Optimal Expectations and Limited Medical Testing:
Evidence from Huntington Disease.” American Economic Review 103 (2): 804–30.
Rabin, Matthew. 1993. “Incorporating Fairness into
Game Theory and Economics.” American Economic
Review 83 (5): 1281–1302.
Rabin, Matthew. 1998. “Psychology and Economics.”
Journal of Economic Literature 36 (1): 11–46.
Rey-Biel, Pedro. 2008. “Inequity Aversion and Team
Incentives.” Scandinavian Journal of Economics 110
(2): 297–320.
Roider, Andreas, and Patrick W. Schmitz. 2012. “Auctions with Anticipated Emotions: Overbidding,
Underbidding, and Optimal Reserve Prices.” Scandinavian Journal of Economics 114 (3): 808–30.
Rosenkranz, Stephanie, and Patrick W. Schmitz. 2007.
“Reserve Prices in Auctions as Reference Points.”
Economic Journal 117 (520): 637–53.
Rothschild, Michael, and Joseph Stiglitz. 1976. “Equilibrium in Competitive Insurance Markets: An Essay
on the Economics of Imperfect Information.” Quarterly Journal of Economics 90 (4): 629–49.
Salant, Yuval, and Ron Siegel. 2013. “Framing in Mechanism Design.” Unpublished.
Sandroni, Alvaro, and Francesco Squintani. 2007.
“Overconfidence, Insurance, and Paternalism.”
American Economic Review 97 (5): 1994–2004.
Sandroni, Alvaro, and Francesco Squintani. 2010.
“Overconfidence and Asymmetric Information: The
Case of Insurance.” Unpublished.
Santos-Pinto, Luís. 2008. “Positive Self-Image and
Incentives in Organisations.” Economic Journal 118
(531): 1315–32.
Schumacher, Heiner. 2012. “Insurance, Self-Control
and Commitment.” Unpublished.
Schweizer, Nikolaus, and Nora Szech. 2013. “Optimal Revelation of Life-Changing Information.”
Unpublished.
Shunda, Nicholas. 2009. “Auctions with a Buy Price:
The Case of Reference-Dependent Preferences.”
Games and Economic Behavior 67 (2): 645–64.
Sliwka, Dirk. 2007. “Trust as a Signal of a Social Norm
and the Hidden Costs of Incentive Schemes.” American Economic Review 97 (3): 999–1012.
Spiegler, Ran. 2006. “The Market for Quacks.” Review
of Economic Studies 73 (4): 1113–31.
Spiegler, Ran. 2011. Bounded Rationality and Industrial Organization. Oxford and New York: Oxford
University Press.
Spinnewijn, Johannes. 2012. “Unemployed but Optimistic: Optimal Insurance Design with Biased
Beliefs.” Unpublished.
Spinnewijn, Johannes. 2013. “Insurance and Perceptions: How to Screen Optimists and Pessimists.” Economic Journal 123 (569): 606–33.
1118
Journal of Economic Literature, Vol. LII (December 2014)
Stoughton, Neal M., Youchang Wu, and Josef Zechner.
2011. “Intermediated Investment Management.”
Journal of Finance 66 (3): 947–80.
Thaler, Richard H., and Shlomo Benartzi. 2004. “Save
More Tomorrow™: Using Behavioral Economics
to Increase Employee Saving.” Journal of Political
Economy 112 (Special Issue 1): S164–87.
Thaler, Richard H., and Cass R. Sunstein. 2003. “Libertarian Paternalism.” American Economic Review 93
(2): 175–79.
Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge:
Improving Decisions about Health, Wealth, and
Happiness. New York: Penguin Books.
Tirole, Jean. 2009. “Cognition and Incomplete Contracts.” American Economic Review 99 (1): 265–94.
Tversky, Amos, and Daniel Kahneman. 1991. “Loss
Aversion in Riskless Choice: A Reference-Dependent Model.” Quarterly Journal of Economics 106
(4): 1039–61.
von Siemens, Ferdinand A. 2011a. “Heterogeneous
Social Preferences, Screening, and Employment Contracts.” Oxford Economic Papers 63 (3):
499–522.
von Siemens, Ferdinand A. 2011b. “Intention-Based
Reciprocity and the Hidden Costs of Control.”
Unpublished.
von Siemens, Ferdinand A. 2012. “Social Preferences,
Sorting, and Competition.” Scandinavian Journal of
Economics 114 (3): 780–807.
Warren, Patrick L., and Daniel H. Wood. 2010. “Will
Governments Fix What Markets Cannot? The Positive Political Economy of Regulation in Markets with
Overconfident Consumers.” Unpublished.
© Copyright 2026 Paperzz